250 : Solving the Impossible: Teaching Robots to See | Gokul NA (Founder,CynLr)

[00:00:00] If you're too sharp, if you're too smart, you wouldn't plunge into this.

[00:00:03] Yeah, I think there is some amount of stupidity involved and I think that is important.

[00:00:07] Gokul is the founder of Synler.

[00:00:09] Deep-tech robotics startup Synler or Cybernetics Laboratory

[00:00:12] is changing the way robots see and work.

[00:00:15] Synler creates smart vision systems that help machines recognize and handle objects, much like humans do.

[00:00:22] How do you iterate in this kind of an environment, right?

[00:00:24] Because it's not like your typical software fail fast and there's a significant cost to getting it wrong.

[00:00:30] You gotta start thinking from, oh, it's going to fail. Am I equipped to tackle it?

[00:00:34] You need to plan your failures.

[00:00:35] So where can people seek help?

[00:00:36] Tech, I think you're on your own. In the US, robotics has a very bad reputation.

[00:00:40] It has not been a very successful investment thus far.

[00:00:42] What do you see as the future of humans?

[00:01:02] Hey Gokul, welcome to the Startup Operator Podcast. Thank you so much for making the time.

[00:01:07] Thanks, thanks Roshan.

[00:01:08] Yeah, it's been a long time coming. I know we had to record a month ago, but I'm happy that it's happening right now.

[00:01:14] And you have a fantastic office, perhaps the coolest office we've seen.

[00:01:18] Possibly, thanks.

[00:01:19] Yeah.

[00:01:20] Yeah, yeah.

[00:01:21] So, I'd like to start at the beginning, right?

[00:01:24] I mean, every business or every startup has this what if, right?

[00:01:28] I mean, what if I can deliver food or groceries in 10 minutes?

[00:01:31] What if I can fit my entire music in my pocket and so on?

[00:01:36] So, what is that what if for SignLo?

[00:01:39] I don't have a clickbait statement to begin with.

[00:01:42] So, for us, it's a very, very long journey, both for Nikhil and I.

[00:01:47] The bug caught me first, but it was not to this clarity that I have today.

[00:01:54] And ours is a very slow frog in the warm water process, right?

[00:01:58] So, this has been a thought process of what's happening with robotics and automation.

[00:02:04] Again, basically, a need for an automation starting all the way from my 11th, 12th around that time.

[00:02:10] And I was picking engineering and then I picked engineering because I wanted to build these machines, especially because I come from a borderline town plus village setup where most of the things were already mechanized and automated.

[00:02:23] But there were these tasks, many of these tasks where you have to do, pick, manipulate, sophisticated tasks or not.

[00:02:30] And labor was very, very short in those places, right?

[00:02:34] That's also truth in India, right?

[00:02:36] It's affordability of automation that's the problem on the other hand, right?

[00:02:40] So, that thought process was there for a long period of time.

[00:02:44] And then even during the engineering phase, some of the clarity between what a prototype and as an engineer, often we have this hobbyist thinking, right?

[00:02:53] I'm good at doing something and somehow I'm going to convert this into an organization and so on.

[00:02:57] But the thought process is very different when you approach from a hobbyist perspective than to a product that you need to make someone else benefit out of and should be usable to someone.

[00:03:06] So, with that background again, was looking at the maturity of the robotics as a technology.

[00:03:13] This is back in 2007, 8, 9 around that time.

[00:03:16] Right.

[00:03:17] One key problem that we noticed also is the first question that I'm thinking is, see, the cars have been there for almost maybe 30 years earlier than what robotic arms were.

[00:03:28] Robotic arms were from 1950s, right?

[00:03:30] First arms that came.

[00:03:31] Why isn't it not percolating and penetrating into a common life and human life, even in industries and businesses or low-level businesses like how a truck and tractor has penetrated into a farm?

[00:03:42] Right. Why haven't they actually penetrated?

[00:03:44] The biggest problem is volume, right?

[00:03:46] And if volume is, why is the volume not clicking in?

[00:03:49] The more it's not generalized enough, the more it's not useful enough for multiple different utilities, it becomes a problem for you to adopt in volumes.

[00:03:57] Right. So, there is a generalization that is lacking.

[00:04:00] Then if we dig a little more deeper, these machines are, these are just simple machines.

[00:04:04] They don't justify the word robot.

[00:04:06] They are just six motors on a link.

[00:04:08] And somehow it has always been an undercurrent of an existing technology back in that time.

[00:04:12] Something market finds as a new tech, they just start coming back and saying, hey, can the robots be better?

[00:04:17] Right.

[00:04:17] Now the fourth phase or third phase is happening.

[00:04:19] Three phases of advance of AI and this is the third phase that's actually happening.

[00:04:24] Where again, we are looking at charge opinion, open AI, and then we are thinking whether this could be extended further into making robots more intelligent.

[00:04:31] Right.

[00:04:32] So those issues were there.

[00:04:35] When, what I realized is if I don't have a general purpose robot system, instead of making it very purpose specific to, if I have, I have an Attenbolt if I have to make, I'll be one, there'll be one specialized robot for it, for welding another robot, for painting another robot.

[00:04:51] And in fact, if you look at the way industry in itself is not able to adopt robots in large volume, only then it will first come, first it has to be luxury market.

[00:04:57] Then it has to come to your consumer market and cost sensitive market.

[00:05:01] Starts with a niche and then gains adoption.

[00:05:02] And then yeah.

[00:05:04] Like your Tesla Roadster first and then now your thing, your Model 3 and so on and so forth.

[00:05:08] I don't remember the names, but the first was the luxury car.

[00:05:11] Right.

[00:05:12] And that's how everything starts.

[00:05:15] The industry which could be profitable, which could offer at a much higher cost in itself was not able to adopt it.

[00:05:20] Because today, if you take 30%, if I take a solution of robot automation, only 30% of the cost goes to the robotic arm.

[00:05:28] Remaining 70% is all about the environment, presenting the parts and so on and so.

[00:05:33] Customization?

[00:05:33] Customization. Heavy customization.

[00:05:35] You have to control your environment, you have to control how your part comes.

[00:05:38] And MM of orientation difference becomes a problem for a robot to go, interact with them, pick them, manipulate them.

[00:05:44] It's a huge problem, right?

[00:05:45] Now imagine, you have to control the presentation of all the parts that comes.

[00:05:49] You have nut and bolt within the space and tolerance of, you know, 0.2 mm.

[00:05:54] Right?

[00:05:55] And size changes, then whole mechanism changes.

[00:05:58] So for some of you realize that, no, I can't go and pick it like this.

[00:06:00] I have to pick it upside down.

[00:06:02] Then the whole mechanism again changes.

[00:06:04] Wow.

[00:06:04] Right?

[00:06:05] With this limitation, how do you automate?

[00:06:07] Human being is obviously a better alternative to, and people might argue, think, because if you go to YouTube and then see those videos in your, you go for, you know, automotive manufacturing, you'll see a lot of robots there.

[00:06:19] Because those are the only videos that are seen.

[00:06:21] Right?

[00:06:21] And often this used to be another question coming from the market saying, hey, but isn't everything to be automated is already automated.

[00:06:28] Right?

[00:06:28] What is, Ilan was claiming about so many robots and so on and so forth.

[00:06:33] There are two things.

[00:06:35] One, if you look at today, who's the largest employer among the organized sector into blue collar labor, right?

[00:06:44] It's essentially manufacturing.

[00:06:46] And automotive manufacturing is one of the largest.

[00:06:49] Construction and agriculture is a little unorganized sector, though they must be the largest.

[00:06:53] Right?

[00:06:54] Among the organized sector, they are the largest.

[00:06:55] So what are these people doing?

[00:06:57] Right?

[00:06:57] If you think about the wages that are paid by organizations like just Ford or $150 billion or $130 billion, organization GM and all of them, they are in like several tens of billions of dollars.

[00:07:08] Right?

[00:07:09] I think Ford is around $7 billion or so.

[00:07:11] So it's just wages alone being paid annually.

[00:07:14] Right?

[00:07:14] So what is it doing?

[00:07:15] What are those many people doing?

[00:07:16] Right?

[00:07:17] On the blue collar shop floors.

[00:07:19] Essentially, what we forgot is mechanization versus robot autonomy.

[00:07:24] Right?

[00:07:25] You still need somebody to operate those machines.

[00:07:28] Right?

[00:07:28] Or move objects from one location to the other location.

[00:07:31] That's what people are doing today.

[00:07:32] Right?

[00:07:33] Or move tools, interact with the tools, assembling an engine.

[00:07:36] Let's say most of those parts are complicated.

[00:07:38] Imagine the number of wires that go into your car, your roof, all your carpeting inside.

[00:07:43] From there to your mobile phone, even this mic and your laptop and all of these are highly unautomitable.

[00:07:50] Because you cannot control them.

[00:07:53] How do you control this wire's presence in that particular angle and orientation?

[00:07:56] Why is that a problem?

[00:07:57] Because these are just simple machines with joysticks.

[00:08:01] You tune your joysticks to move it to a particular location, record that position, record to the other.

[00:08:06] It will just repeat the moment without a deviation from that moment within like possibly point to mm to point to mm of repeatability.

[00:08:15] Right?

[00:08:15] But it's completely dumb.

[00:08:17] They don't understand what's happening in the environment and so on.

[00:08:20] So, fine, putting a camera and then making an intelligent system, making it intelligent by writing algorithms behind it.

[00:08:28] That's where the key genesis point.

[00:08:30] So, I'm boiling it all down to this particular point and then finally, okay, how do I make it adaptable?

[00:08:34] Then you need a sensor that's more holistic to give them a feedback about what's happening in the environment.

[00:08:39] Now, if you put a holistic sensor, vision is the largest holistic sensor where you can see a whole area of information, not like a point information.

[00:08:47] Your touch is a point information.

[00:08:49] Right?

[00:08:49] Vision is like million pixels.

[00:08:51] You're seeing all of them.

[00:08:52] You can see depth.

[00:08:52] You can see a lot of information.

[00:08:55] Unfortunately, we have underestimated the approach of using vision.

[00:08:59] We always keep associating that with, you know, comparing images.

[00:09:04] We think that, you know, if I want to understand an object, if I want to find, let's say there is a pen in this room.

[00:09:09] It's my pen.

[00:09:10] It's my favorite pen.

[00:09:11] Let's say I want to go search for the pen.

[00:09:12] Again, the assumption that we think how our brain works is it has an image of the pen in its head and goes about searching in the environment.

[00:09:19] Right?

[00:09:20] And sorry, starts comparing them and then starts figuring that out.

[00:09:24] What are we comparing?

[00:09:25] We are thinking colors.

[00:09:26] But the color for a reflective mirror finished object or a transparent object, what is color?

[00:09:30] It's the whole reflection of the environment in itself that becomes very complex for the system to understand.

[00:09:35] Right?

[00:09:36] So this is underestimated.

[00:09:38] We, human beings have a tendency to usually underestimate some of these processes that we don't understand.

[00:09:45] While the progress has been phenomenal when it comes to language and language processing, it has been very abysmal when it comes to vision and vision oriented processing.

[00:09:54] Right?

[00:09:55] That's because this is the only process that is too natural and instinctive to us.

[00:09:59] Like to talk, we went to school.

[00:10:01] Right?

[00:10:01] To read and write, we went to school.

[00:10:03] Somebody taught us how to eat.

[00:10:04] Somebody taught us how to dress.

[00:10:05] Somebody taught us possibly held our hand and then made us walk and so on and so forth.

[00:10:09] Right?

[00:10:10] But when it comes to vision, you just opened your eyes.

[00:10:12] It happened.

[00:10:12] You didn't read a book how to improve your vision.

[00:10:14] None of those.

[00:10:14] Right?

[00:10:15] So it just happened.

[00:10:16] Right?

[00:10:18] And the amount of underestimation of processing that's happening in your brain.

[00:10:21] Right?

[00:10:22] Given time, 55% of your brain is occupied for visual processing.

[00:10:26] Wow.

[00:10:26] That's why you close your eyes when you are smelling a rose or let's say, or anything good smelling for the case.

[00:10:32] Right?

[00:10:32] Or when you want to listen to music or when you're thinking deep or when you're eating an ice cream, the first instinct is to, right?

[00:10:39] So the eye goes closing.

[00:10:41] Right?

[00:10:41] So that is a key factor that we are underestimating.

[00:10:46] Why?

[00:10:47] Because brain has to relieve its processing so that the other senses are in a heightened state.

[00:10:52] Because the vision consumes, how are we simplifying it?

[00:10:56] How are we oversimplifying it?

[00:10:58] Have we understood the working of the vision in itself?

[00:11:00] So those are all large gaps that we realized.

[00:11:04] And when I was getting into, so there was a huge gap in the vision and the vision is so underestimated.

[00:11:09] And that's the, with that notion, I joined National Instruments as a company right out of my college.

[00:11:17] So with vision, focusing on the vision.

[00:11:20] So both Nikhil and I met there in 2011 at National Instruments.

[00:11:25] And everybody starts at that particular role and then diversifies into different roles later.

[00:11:30] And I moved on to become a mission vision specialist, interfacing between the R&D and application engineering team in National Instruments.

[00:11:38] One of the again plaguing problem that the industry had is if you attempt 10 problems, only three would commercially succeed.

[00:11:43] Right? That's the mission vision's reputation.

[00:11:45] Customers would say, hey, if you're solving something with vision, please don't.

[00:11:50] If you have something that you could do with mechanical wear, I would rather control my environment than to put a vision on top of my robot and then having one person stand all the time to see whether it is going to work this time or next time or when it's not going to work.

[00:12:01] Right?

[00:12:02] That's again also boiled down when we did the soul searching why this is happening.

[00:12:06] And it went to the fact that we are again assuming manipulation from identification.

[00:12:12] Hmm.

[00:12:13] What is the foundation fallacy?

[00:12:15] This template picture, if you're not able to match, you can't go act on this object.

[00:12:19] Right?

[00:12:19] If I pull out a random object of my, like, let's say my key chain that you've never seen before,

[00:12:24] or a crushed paper and I give it to you.

[00:12:26] It's a completely random object, right?

[00:12:28] You would have never had the mental picture of what it is.

[00:12:30] There's no template.

[00:12:30] There's no template.

[00:12:31] Yet you will be able to go pick it.

[00:12:32] You're able to use your vision, guide yourself, your faculties, which is hand and whatnot, focus on that object, go pick it, which a baby can.

[00:12:43] Baby may not know you're feeding a pen, right?

[00:12:45] You're giving it a pen.

[00:12:46] He'll pick it, put it in his mouth, throw it out.

[00:12:48] Right.

[00:12:48] So these are instinctual aspects of how it interprets the world, which is absent in vision.

[00:12:53] Right.

[00:12:53] What we call a sentience before intelligence.

[00:12:56] I mean, people think sentience is emotions and then a higher elevated stage of intelligence, but it's more about the more primal instincts that the system has.

[00:13:03] Like if you touch a fire, your body understands.

[00:13:06] Sometimes if you, if an object, I mean, may not be like snake, but something else that we instinctively babies do react to.

[00:13:15] And those are structures that are built in your brain, which is used to learn and interpret the world.

[00:13:20] That is a large gap that the machines have today to be able to interact with the physical world, extract the intelligence about or not intelligence, the information about the world and the objects around.

[00:13:34] And, and those missing pieces is what we realized is, is, is, is distinct is a distinction for a human being to be able to manipulate really well.

[00:13:43] And identification is a unintended consequence of the brain wanting to manipulate the world around it.

[00:13:51] Right.

[00:13:52] Right.

[00:13:52] It's not that we used identification to be able to become manipulation savvy.

[00:13:56] We did manipulation of space, either navigating or whatnot.

[00:14:01] Then identification became a, because there are animals which can, cannot remember yet.

[00:14:06] Yet they can go act on the, to the extent that we can.

[00:14:10] Right.

[00:14:10] But they, it can act on items around it, the environment around it.

[00:14:14] Right.

[00:14:14] So this gap is what we noticed.

[00:14:16] And then we went out in 2015, we left national instruments, both of us.

[00:14:22] Nikhil was specializing more into the business side of, more into the sales side of the, and our con management at national instruments.

[00:14:29] And while I was focusing on the vision side of things and building these architectures and teams and so on.

[00:14:34] Then in 2015, we stepped out to test our thesis.

[00:14:37] So we took, with the blessings of national instruments, we, we took some 30 different failed projects and problems in the part of the past.

[00:14:46] And then after we had successfully, you know, cleared all of them, we had a hundred percent success rate in a space where you had 70 to 30% ratio of success failure.

[00:14:56] We just took this as a precedence to go race and then productize this because we were trying, we were dependent, hyper dependent on the off the shelf components.

[00:15:04] It was not packaged and form factored into the way where the vision in itself, the camera in itself, you don't have an off the shelf camera that can give you all these abilities.

[00:15:12] It was limiting.

[00:15:13] So we started building, we started looking at what does the customer want?

[00:15:16] He wants a complete automation solution, which needs a plethora of different manipulation capabilities, sensors, grippers and vision.

[00:15:22] And vision was the foundation that you had to begin with. So we started with building that from the hardware in itself.

[00:15:27] And from 2019 onwards, we have been, that has been the journey.

[00:15:30] Fantastic.

[00:15:32] So if I understand correctly, machines have become very good at very specific purposes, right?

[00:15:38] But what you're trying to do is make it adaptable enough for a wide variety of things that it can do, wide variety of tasks and operations and so on.

[00:15:44] Right. It seems like such a fundamental thing, right? Why do you think there hasn't been much development on this for the last 40, 50 years?

[00:15:51] Or, or do you think that people have taken a different view of, you know, what they could get the machines and robots to do?

[00:15:58] I think there are multiple parameters and factors when there are, yeah, there are several parameters, right?

[00:16:06] So one, how the industry approaches in itself, right?

[00:16:10] There's an industry angle to it.

[00:16:11] I mean, everything boils down to people's perception and different departments understanding it in a different way, right?

[00:16:15] Of businesses understanding it in a different way.

[00:16:18] There is a fallacy of confusing with integration and developing something fundamental.

[00:16:24] Robotics had always been seen as an application domain than a domain that needs fundamental research, right?

[00:16:32] We are not noticing that it's a component level problem.

[00:16:35] We are not realizing that it's a technology level problem.

[00:16:38] Something has been not working for more than 40 to 60 years now that you're attempting.

[00:16:43] The first camera on top of a robotica was tried in the 70s, right?

[00:16:46] Unlock camera, you don't have unlock processing, you don't have digital computers and none of this.

[00:16:51] I'm not able to find this video again, but it was some MIT lab attempting that back in the time, right?

[00:16:56] What's missing is we always thought as it's an outcome of disparate set of technologies that is to, you could put them together and make it work.

[00:17:05] Right?

[00:17:05] We didn't think it needs its own dedicated set of technology by itself.

[00:17:10] That's a mistake.

[00:17:11] Okay.

[00:17:11] And it has always been thought as somebody's incapability of engineering with all the existing tech.

[00:17:19] It's possible and then, and, and foundationally vision is very, very illusionary, elusive.

[00:17:25] Right?

[00:17:27] So that's why it's elusive also, right?

[00:17:29] So we, we, we do not understand how the process of vision evolves.

[00:17:34] So we underestimated, kept underestimating, kept underestimating.

[00:17:37] It goes on even now.

[00:17:39] The problem is the moment you see a picture or an image or a video, we, we are able to understand it really well.

[00:17:45] Right?

[00:17:46] We assume that that's what your brain is also saying.

[00:17:49] There is this personification issue.

[00:17:51] Anything that moves like a human being, we think it has the intelligence like a human being.

[00:17:54] The moment we see those pictures, we think that we understand.

[00:17:57] So the machine must be understanding.

[00:17:58] Oh, it's able to break.

[00:18:00] So if I do some kind of filter, you're able to see the edges and contours.

[00:18:03] Oh, it's able to extract those edges.

[00:18:05] I'm able to recognize that it's a human being.

[00:18:07] So which means the machine must be, you know, interpreting it.

[00:18:09] It's possible for, to make the machine understand that it's a human being.

[00:18:12] So those fallacies makes us, you know, underestimate the nuances that are there in the, in the, in solving this problem.

[00:18:21] And we're not starting from the problem.

[00:18:23] Many times it's engineers who have been very savvy with particular tools and toolkits.

[00:18:28] And, and when I say toolkits, even subjects and disciplines, right?

[00:18:31] They're coming to solve this from that front.

[00:18:34] They're not coming from the problem.

[00:18:36] If an object is so hard and they have been attempting for 40 years, it's something about the object is a problem, isn't it?

[00:18:41] It need not be the tech.

[00:18:42] So what am I not understanding about the problem?

[00:18:45] Right?

[00:18:45] This is one.

[00:18:46] This is on the technology side.

[00:18:48] On the business side, on the other hand, it has always been an extension of another industry.

[00:18:54] It's an undercurrent, technology undercurrent, right?

[00:18:57] Okay.

[00:18:58] Where, oh, this should be extension of some automation of so-and-so.

[00:19:01] This should be an extension of CNC missions.

[00:19:03] Or this should be extension of, you know, some process production automation.

[00:19:07] And so they, that again, underestimation was a, was a, was a major problem in, in the way they, they try to commoditize, hyper commoditize very quickly.

[00:19:18] Not looking at this as a vertical integration problem.

[00:19:21] Got it.

[00:19:22] Right?

[00:19:22] Right.

[00:19:23] And you're borrowing, you are trying to solve the proofs of market through borrowed components from, from different, different, you know, components of different products, which were built purpose-built for some other product.

[00:19:36] And you're trying to, you know, integrate it all together to make it fit into this fashion.

[00:19:40] Right?

[00:19:40] So these are the key issues, why this problem has been eluding the market for a long period of time.

[00:19:48] So, right.

[00:19:49] The end impact of what you're building seems to be like a universal factory, right?

[00:19:52] I mean, I don't need a different PC or a Mac to code a consumer software versus an enterprise SaaS software.

[00:19:59] Right?

[00:20:00] But I certainly need a different set of machine components to build object A versus object B.

[00:20:06] Right?

[00:20:07] But if you build machines to the level that they're extremely adaptable, then theoretically, then I just need a bunch of these machines and I can produce anything with that.

[00:20:16] Yeah.

[00:20:17] Yeah.

[00:20:18] A bit of an insight into this is universal factory is what we imagine could become.

[00:20:22] Right?

[00:20:23] That's, that's a visualization that we have or the vision that we have towards what this technology can give birth to.

[00:20:31] Right?

[00:20:32] On the other hand, the foundation will come from an object computer principle.

[00:20:35] Right?

[00:20:36] What we, which is basically a borrowed imagination.

[00:20:39] Yeah.

[00:20:39] It's a borrowed imagination.

[00:20:42] It's an analogical, you can think of an analogical process of how electronics evolved and compute platforms evolved.

[00:20:47] Right?

[00:20:48] 1960s and 70s.

[00:20:49] Even before that, the automation was there for a long period of time.

[00:20:52] People used to do the computing through, you know, gears and even before that was pneumatic and so on and so forth.

[00:20:56] Right?

[00:20:56] And vacuum tubes came.

[00:20:57] Right?

[00:20:58] Then people started doing punching cards as inputs to those systems and things.

[00:21:01] But every point of time, whenever there is a customization, people used to have a lot of different machines, custom built and customs or queues and so on and so forth.

[00:21:08] Right?

[00:21:08] They were not specific computers built.

[00:21:10] There are computers that were built for specific processing.

[00:21:13] Right?

[00:21:14] In fact, you would have famously seen this picture where in 1980s, there are so many devices on your desktop, which has just shrunk down now to a computer and now to a mobile phone altogether.

[00:21:27] So, which all became…

[00:21:28] This revolution and transformation happened when they realized that hardware has to be standardized.

[00:21:34] Software is still not standardized.

[00:21:36] Right?

[00:21:37] What you could do is you can have a Swiss Army knife of a device which is equipped with…

[00:21:43] Am I 24 bar 7 using the webcam there?

[00:21:45] No.

[00:21:46] Am I using the microphone 24 bar 7?

[00:21:47] No.

[00:21:48] Am I HAWground.

[00:21:50] The program conditions …

[00:21:56] No.

[00:22:00] You've got…

[00:22:01] It's a...

[00:22:05] …

[00:22:11] Well, that that is...

[00:22:14] That's the idea that the next time...

[00:22:16] important won't be tap on OK.

[00:22:17] It's a volume problem.

[00:22:19] Utility to be universal

[00:22:20] is how you create

[00:22:21] an under denominator

[00:22:22] or the common denominator

[00:22:24] of problems.

[00:22:25] If you don't do that,

[00:22:26] you can't productize.

[00:22:28] So what laptops brought

[00:22:30] in terms of hardware standardization

[00:22:32] is what we try to bring

[00:22:33] in terms of robots today.

[00:22:35] It could be,

[00:22:36] again,

[00:22:36] it can go into the thought process

[00:22:37] of what all does it need

[00:22:39] as a bare minimum standardization

[00:22:40] that we can get into,

[00:22:41] which is modular enough

[00:22:44] where today I could use it

[00:22:46] for nut and bolt,

[00:22:46] tomorrow I could use it

[00:22:47] for wiring something.

[00:22:48] Day after tomorrow,

[00:22:49] I could use it possibly

[00:22:49] for stitching a cloth.

[00:22:50] Today we can't.

[00:22:51] It's a very complicated problem

[00:22:52] to solve.

[00:22:53] But eventual goal

[00:22:54] is to get there,

[00:22:55] that you have one

[00:22:55] underlying hardware.

[00:22:56] The moment you have

[00:22:57] one underlying hardware,

[00:22:58] making it do something else

[00:23:00] is virtual.

[00:23:00] It's on your software.

[00:23:02] Front, right?

[00:23:04] Now, that's the journey

[00:23:05] that we want to get into.

[00:23:06] Now, if you're able to do that,

[00:23:09] why is a factory,

[00:23:11] which is,

[00:23:11] now if you think about

[00:23:13] to make a car,

[00:23:15] to design a car

[00:23:16] or design a factory,

[00:23:17] which one consumes

[00:23:17] the maximum amount

[00:23:18] of money for a customer today?

[00:23:21] Factory.

[00:23:21] Factory.

[00:23:22] Shouldn't that be

[00:23:23] the product

[00:23:24] that he actually owns

[00:23:25] than the producer

[00:23:26] in itself?

[00:23:27] But he is bottlenecked

[00:23:29] to buy the product

[00:23:29] in itself, right?

[00:23:31] Because this infrastructure

[00:23:32] that you have produced

[00:23:32] is specific

[00:23:33] to only that product.

[00:23:34] So, which comes

[00:23:34] with such a large,

[00:23:35] huge amount

[00:23:36] of business risk

[00:23:37] that someone else

[00:23:38] evolves and then

[00:23:39] has an advantage

[00:23:40] to look at his mistakes

[00:23:41] and then evolve

[00:23:42] into a new business.

[00:23:43] He is not able

[00:23:44] to compete along

[00:23:44] with them

[00:23:45] because I have to

[00:23:45] make a recovery

[00:23:46] of this investment,

[00:23:47] capital investment

[00:23:48] that has gone

[00:23:48] into this business.

[00:23:50] Right.

[00:23:50] We could say that as

[00:23:52] industry keeps thinking

[00:23:53] that as a constraint

[00:23:53] to solve,

[00:23:54] we are thinking

[00:23:54] that could be a problem

[00:23:56] that is solvable.

[00:23:58] The constraint is,

[00:23:59] oh, you can't have

[00:24:00] non-specialized machinery.

[00:24:01] The machinery is what

[00:24:02] it boils down to the machinery

[00:24:03] and the process

[00:24:04] to operate the machinery.

[00:24:05] If I could generalize

[00:24:06] the machinery,

[00:24:07] the machinery can adapt

[00:24:08] itself to various

[00:24:08] different tasks.

[00:24:10] the sequence of whatever

[00:24:12] I did here could be

[00:24:13] for another microphone.

[00:24:15] Right.

[00:24:16] That someone else

[00:24:17] sitting somewhere else

[00:24:17] could actually,

[00:24:18] you know,

[00:24:20] teach and train

[00:24:20] that robot

[00:24:21] and this robot here

[00:24:22] will transform itself

[00:24:23] to immediately

[00:24:23] do something else,

[00:24:25] some other recipe

[00:24:26] of task.

[00:24:27] Right.

[00:24:27] Right.

[00:24:27] So this is,

[00:24:29] when this happens,

[00:24:30] the factory could

[00:24:31] scale itself

[00:24:32] to a variety

[00:24:33] of different,

[00:24:34] you know,

[00:24:35] product

[00:24:35] because product

[00:24:36] is nothing but

[00:24:37] same aluminum

[00:24:38] becomes this,

[00:24:39] the stand

[00:24:40] in the same

[00:24:40] aluminum

[00:24:41] becomes that robot.

[00:24:42] Right.

[00:24:42] Or a car.

[00:24:44] Right.

[00:24:44] Or another stand.

[00:24:45] Right.

[00:24:45] The difference is

[00:24:46] how the set of

[00:24:48] procedures that you did

[00:24:49] to change them.

[00:24:50] Right.

[00:24:51] So rather than,

[00:24:52] you know,

[00:24:52] optimize the input

[00:24:53] parameters,

[00:24:54] make it cheaper,

[00:24:54] faster,

[00:24:54] better,

[00:24:55] can I increase

[00:24:56] the outcomes,

[00:24:57] vary the outcomes

[00:24:58] in such a way

[00:24:59] that with the same

[00:24:59] setup,

[00:25:00] I mean,

[00:25:00] I can do

[00:25:00] many different things.

[00:25:02] And so the ROI

[00:25:03] on that is...

[00:25:04] is a matter of time

[00:25:05] and investment

[00:25:05] back into the

[00:25:06] component level.

[00:25:08] To the components

[00:25:08] that goes into

[00:25:09] the arms

[00:25:09] to make them

[00:25:10] more and more

[00:25:10] efficient and so

[00:25:11] on and so forth.

[00:25:12] Right.

[00:25:12] So that will be

[00:25:13] the Moore's law

[00:25:13] for electronics

[00:25:14] that we can.

[00:25:15] I don't know,

[00:25:15] I'm just guessing

[00:25:16] that there will be

[00:25:17] a law like Moore's

[00:25:18] law that will come

[00:25:19] for us.

[00:25:20] When you start

[00:25:21] with such a,

[00:25:22] like a huge

[00:25:23] ambition,

[00:25:24] right,

[00:25:24] I mean,

[00:25:25] it's really like

[00:25:26] out there.

[00:25:28] What does day one

[00:25:29] look like?

[00:25:30] I mean,

[00:25:30] what is the first

[00:25:30] problem you pick

[00:25:31] to solve?

[00:25:33] Right.

[00:25:33] And how do you

[00:25:34] get your team

[00:25:34] to kind of focus

[00:25:35] on that with

[00:25:37] drawing a straight

[00:25:37] line to that

[00:25:38] end vision

[00:25:38] that you have?

[00:25:40] Yeah,

[00:25:41] I think there is

[00:25:41] some amount

[00:25:42] of stupidity

[00:25:42] involved

[00:25:45] in thinking

[00:25:46] that you could

[00:25:46] solve it.

[00:25:47] And I think

[00:25:48] that is important.

[00:25:49] If you're too

[00:25:49] sharp,

[00:25:50] if you're too

[00:25:50] smart,

[00:25:51] you wouldn't

[00:25:52] plunge into

[00:25:52] this,

[00:25:53] right?

[00:25:53] That is some

[00:25:54] amount of

[00:25:54] element of

[00:25:55] hope.

[00:25:55] And plus,

[00:25:55] the question

[00:25:56] of day one,

[00:25:57] I think,

[00:25:58] like I said,

[00:25:58] it's a,

[00:25:59] for me at least,

[00:26:00] it was a gradual

[00:26:02] evolution of

[00:26:02] industry reflecting

[00:26:03] on you,

[00:26:04] you reflecting on

[00:26:05] industry,

[00:26:05] right?

[00:26:06] So it's been

[00:26:07] a gradual

[00:26:07] growth,

[00:26:08] right?

[00:26:08] And then we

[00:26:09] had a long-term

[00:26:10] picture.

[00:26:10] We do not,

[00:26:11] that's like,

[00:26:11] you know,

[00:26:12] you have a peak,

[00:26:13] something that

[00:26:13] you want to,

[00:26:14] the K2 that you

[00:26:15] want to climb,

[00:26:16] you have no

[00:26:17] path laid.

[00:26:18] It's like a

[00:26:18] forest,

[00:26:18] right?

[00:26:19] So we have to

[00:26:19] group ourselves,

[00:26:21] find a bunch of

[00:26:22] people who are

[00:26:22] exploratory,

[00:26:23] go a set of

[00:26:23] milestones and

[00:26:24] then come back

[00:26:25] again,

[00:26:25] regroup,

[00:26:26] say which is the

[00:26:26] path that you

[00:26:27] could move

[00:26:27] forward and move

[00:26:28] forward with

[00:26:28] keeping that in

[00:26:29] the focus.

[00:26:30] Because that's a

[00:26:30] long,

[00:26:31] both in time

[00:26:32] and the way and

[00:26:33] the journey is

[00:26:35] unpredictable and

[00:26:35] very long.

[00:26:37] The first

[00:26:37] structure is for

[00:26:38] us to understand

[00:26:39] what is the

[00:26:40] first milestone

[00:26:40] that we can

[00:26:41] achieve that is

[00:26:41] visible to us

[00:26:43] from the

[00:26:43] viewpoint or

[00:26:44] vantage point of

[00:26:44] where we are,

[00:26:45] right?

[00:26:46] So blending

[00:26:48] along with the

[00:26:49] industry is a

[00:26:50] primary,

[00:26:51] you know,

[00:26:52] direction setter

[00:26:52] for you.

[00:26:53] And we have to

[00:26:54] start,

[00:26:54] what is that

[00:26:54] that we can

[00:26:55] actually start

[00:26:56] today that the

[00:26:57] customer could be

[00:26:57] profitable with,

[00:26:59] right?

[00:27:00] And then start

[00:27:01] evolving and then

[00:27:01] moving further and

[00:27:02] further away with,

[00:27:03] right?

[00:27:03] So of course,

[00:27:05] I would not have

[00:27:06] a maturity of the

[00:27:06] technology on the

[00:27:07] day one because

[00:27:08] it's first of all

[00:27:09] foundational technology

[00:27:10] which means proving

[00:27:11] it to a product form

[00:27:12] factor and it still

[00:27:13] will take a lot of

[00:27:13] time.

[00:27:14] And you don't have

[00:27:15] support of an

[00:27:16] ecosystem,

[00:27:16] your price points

[00:27:17] are not going to be

[00:27:18] mature,

[00:27:18] your talent,

[00:27:21] the level of talent

[00:27:22] that you have and

[00:27:23] the quantum of

[00:27:24] talent,

[00:27:24] quantity of talent

[00:27:25] that you have will

[00:27:26] also be very

[00:27:27] limited for it to

[00:27:28] scale and so on

[00:27:29] so forth.

[00:27:30] So often the

[00:27:31] interpretations of

[00:27:31] commoditized market

[00:27:32] will not work out.

[00:27:34] if you take the

[00:27:36] template models of

[00:27:37] how a Swiggy or

[00:27:38] a Ola or such

[00:27:40] businesses which are

[00:27:42] business model

[00:27:43] geniuses which

[00:27:45] have,

[00:27:46] which learn to

[00:27:47] leverage the

[00:27:47] infrastructure maturity

[00:27:49] that is there,

[00:27:49] right?

[00:27:50] For us, we have to

[00:27:50] start with infrastructure

[00:27:51] in itself.

[00:27:51] That's the first thing

[00:27:52] that we stare at

[00:27:53] as a challenge,

[00:27:54] right?

[00:27:55] Second,

[00:27:55] industry will not

[00:27:56] provide you with

[00:27:57] talent because this

[00:27:58] industry is non-existent

[00:27:59] today and what you're

[00:28:00] building is not a

[00:28:01] company, you have to

[00:28:01] build an industry.

[00:28:02] industry.

[00:28:02] If you don't have

[00:28:03] that consciously in

[00:28:04] your head, then you

[00:28:05] are underestimating the

[00:28:06] problem, right?

[00:28:09] So with all these

[00:28:11] parameters to play,

[00:28:12] you start looking at

[00:28:13] where all my

[00:28:14] advantages and lead

[00:28:15] way would be,

[00:28:17] right?

[00:28:18] And you have to

[00:28:18] search for those.

[00:28:19] It's a lot of

[00:28:20] experimental

[00:28:22] progression that is

[00:28:23] involved in this.

[00:28:24] So clearly,

[00:28:25] we are staring at a

[00:28:26] lot of unknowns.

[00:28:27] Only thing for us to

[00:28:28] see is where is the

[00:28:30] first hook for us to

[00:28:31] hang on to progress

[00:28:32] further, right?

[00:28:33] Customer is one,

[00:28:34] vendor is the other.

[00:28:35] So you have to

[00:28:36] balance between them.

[00:28:38] Strongly,

[00:28:38] investor would be the

[00:28:39] third.

[00:28:40] Investor would react to

[00:28:41] these hooks being

[00:28:41] available.

[00:28:42] That gives them at

[00:28:43] least a modicum of

[00:28:44] trust to say,

[00:28:45] hey, you could move

[00:28:46] forward.

[00:28:46] There is something

[00:28:47] that we can go for

[00:28:48] the next step and

[00:28:48] the next step.

[00:28:49] So I think that's

[00:28:50] how this market,

[00:28:50] even for a business

[00:28:52] model innovation,

[00:28:53] it's the same case.

[00:28:54] We have to start with

[00:28:55] infrastructure

[00:28:55] innovation in itself.

[00:28:56] So it's even more

[00:28:58] harder, even more

[00:28:58] riskier.

[00:29:00] So, and then we

[00:29:02] keep, as we keep

[00:29:03] grinding, an industry

[00:29:04] starts slowly forming

[00:29:05] around us.

[00:29:07] And the cycle thus

[00:29:09] far, that's how it

[00:29:10] has happened.

[00:29:10] So you find a

[00:29:11] manufacturing service

[00:29:12] that you can offer,

[00:29:13] productize that, and

[00:29:15] then build that

[00:29:15] capability further up,

[00:29:17] right?

[00:29:17] Essentially more than

[00:29:18] the manufacturing

[00:29:18] service.

[00:29:19] I wouldn't put myself

[00:29:20] in a confidence

[00:29:21] No, you limit the

[00:29:22] scope of problems,

[00:29:23] right?

[00:29:23] And then look for

[00:29:25] people who are going

[00:29:25] to pay for you to

[00:29:26] solve this problem.

[00:29:27] That's right.

[00:29:28] Build that thing for

[00:29:29] them and then develop

[00:29:30] the technology and build

[00:29:31] on that innovation to

[00:29:32] make it, you know,

[00:29:33] sufficiently productized

[00:29:34] that you can now take

[00:29:35] that to someone else,

[00:29:37] their peer perhaps,

[00:29:38] and say that,

[00:29:38] look, we've solved

[00:29:39] this problem.

[00:29:39] There is one mistake

[00:29:40] that typically everybody

[00:29:40] does.

[00:29:42] Many system integrators

[00:29:43] who wanted to be a

[00:29:44] product company ended

[00:29:45] up being only a system

[00:29:46] integrator.

[00:29:47] Because you think

[00:29:47] that if you do it

[00:29:48] for one customer,

[00:29:50] it will magically,

[00:29:51] the product versus

[00:29:52] system integration

[00:29:53] thinking is confused

[00:29:54] in many cases,

[00:29:55] right?

[00:29:57] Now, but a step

[00:29:58] back onto your

[00:30:00] point, just resonating

[00:30:01] more onto your

[00:30:01] point.

[00:30:02] It's true, we need

[00:30:04] a customer, but the

[00:30:04] customer is not going

[00:30:05] to define the product

[00:30:06] for you.

[00:30:07] What do you have

[00:30:07] as a technology?

[00:30:08] You have a know-how

[00:30:09] that you have invented

[00:30:10] saying that if you

[00:30:10] have a camera like

[00:30:11] this, if you have a

[00:30:11] robot like this, if it

[00:30:12] has these capabilities,

[00:30:13] it has these features,

[00:30:14] you would be able to

[00:30:15] solve the particular

[00:30:16] problem.

[00:30:16] But would you be able

[00:30:17] to fit into the

[00:30:17] price point?

[00:30:18] Would you be able to

[00:30:19] fit into the

[00:30:20] necessities of

[00:30:21] the customer,

[00:30:22] their processes,

[00:30:23] and then how it

[00:30:24] interacts with the

[00:30:25] rest of the other

[00:30:25] components in their

[00:30:26] industry, right?

[00:30:27] Often what you have

[00:30:29] to realize is you

[00:30:30] don't have the

[00:30:31] convenience of B2C

[00:30:32] where you yourself,

[00:30:33] in itself is complex,

[00:30:34] right?

[00:30:35] Because you yourself

[00:30:36] being a customer,

[00:30:38] you don't have that

[00:30:39] luxury here.

[00:30:40] Your customer,

[00:30:42] your problem is not

[00:30:43] that robot's

[00:30:45] incapability to put

[00:30:45] two parts together.

[00:30:46] Your problem is

[00:30:47] that business.

[00:30:49] If you don't

[00:30:49] understand whether

[00:30:50] the business can

[00:30:51] be profitable

[00:30:51] around this,

[00:30:52] then you don't

[00:30:52] have an organization

[00:30:53] around this,

[00:30:54] you don't have a

[00:30:55] possibility of a

[00:30:56] startup or a

[00:30:57] company around

[00:30:57] this, right?

[00:30:59] What you're working

[00:30:59] on is the business

[00:31:00] problem there.

[00:31:01] You have to trace

[00:31:02] that back, right?

[00:31:04] And then you need

[00:31:05] to take, if you ask

[00:31:06] the customers, they'll

[00:31:07] always tell you

[00:31:07] faster horses.

[00:31:09] Yeah.

[00:31:09] Ford statement,

[00:31:11] right?

[00:31:13] That's because they

[00:31:14] think their solutions

[00:31:15] through what is

[00:31:16] available, what they

[00:31:17] are sure of.

[00:31:18] You're taking a

[00:31:18] technology that is

[00:31:19] not yet proven,

[00:31:20] right?

[00:31:21] The only way by

[00:31:22] which they are

[00:31:23] going to give it to

[00:31:23] you is when you

[00:31:24] own the risk of

[00:31:25] his final utility,

[00:31:26] which is the

[00:31:26] whole solution.

[00:31:28] He's not going to

[00:31:29] bite on to saying

[00:31:30] that, oh, he's

[00:31:31] built a new tech, I

[00:31:32] hope that this is

[00:31:32] going to work.

[00:31:33] I'm very, very

[00:31:34] confident that this

[00:31:34] is going to work.

[00:31:35] So, because his

[00:31:37] neck is at the end

[00:31:38] of the day within

[00:31:38] his organization,

[00:31:39] it's an organization,

[00:31:39] it's not a consumer

[00:31:40] market, right?

[00:31:41] Where he can take

[00:31:41] his own calls,

[00:31:43] right?

[00:31:43] He's answerable to so

[00:31:44] many other things

[00:31:45] and then it could

[00:31:45] cripple the cycle

[00:31:46] of the organization,

[00:31:47] operational cycle of

[00:31:48] the organization and

[00:31:48] so on and so forth.

[00:31:49] So, the system is

[00:31:50] built very strongly

[00:31:51] to ensure that

[00:31:53] mistakes don't

[00:31:53] happen.

[00:31:54] So, the only way

[00:31:55] by which you could

[00:31:55] do is you need to

[00:31:56] take the risk.

[00:31:58] You need to borrow

[00:31:58] the risk.

[00:31:59] The more you mitigate

[00:32:00] the risk for the

[00:32:01] decision maker,

[00:32:02] the sooner he's

[00:32:03] going to, that's

[00:32:03] the reason why you

[00:32:04] have to go with

[00:32:04] a solution entirely.

[00:32:06] And you have to

[00:32:06] prove it to him

[00:32:07] how it will work.

[00:32:08] Then you have the

[00:32:09] next phase of work

[00:32:10] where you have to

[00:32:11] automate this through

[00:32:13] the infrastructure like

[00:32:13] system integrators

[00:32:15] and other suppliers

[00:32:16] and distributors

[00:32:16] that you might have

[00:32:17] as a business chain

[00:32:18] to take it forward.

[00:32:20] And for that,

[00:32:20] you have to further

[00:32:21] productize your

[00:32:22] training,

[00:32:23] you productize your

[00:32:23] support,

[00:32:24] productize your

[00:32:24] what do I mean by

[00:32:25] support and

[00:32:25] productization?

[00:32:26] It should be the same.

[00:32:27] You need to,

[00:32:28] to a certain extent,

[00:32:29] templatize it.

[00:32:30] Otherwise, you can't

[00:32:30] scale.

[00:32:32] Right?

[00:32:32] The scale thinking

[00:32:33] is what is missing

[00:32:34] often when they go

[00:32:34] about a solution.

[00:32:35] Many do figure it

[00:32:36] out.

[00:32:37] Right?

[00:32:37] Whether they

[00:32:38] articulate it this way

[00:32:38] or not, they

[00:32:39] understand this and

[00:32:41] they go with

[00:32:41] solutions.

[00:32:42] In this industry,

[00:32:43] predominantly,

[00:32:43] they go with

[00:32:43] solutions.

[00:32:43] What happens is

[00:32:45] how do you

[00:32:45] productize it?

[00:32:46] So that's where

[00:32:47] the slip happens

[00:32:48] often.

[00:32:49] They leave that

[00:32:50] to market to

[00:32:50] dictate it for

[00:32:51] you.

[00:32:52] You need to

[00:32:52] constantly make

[00:32:53] effort.

[00:32:55] Unfortunately,

[00:32:56] this model is

[00:32:57] not very VC

[00:32:58] savvy or VC

[00:33:00] exposed.

[00:33:01] And VCs are

[00:33:02] not exposed to

[00:33:02] this model.

[00:33:03] To make it work,

[00:33:04] it's a long

[00:33:04] gestation thing.

[00:33:05] Neither their

[00:33:06] fund thesis

[00:33:06] allows you to

[00:33:07] do that because

[00:33:07] it consumes a

[00:33:08] lot of cost.

[00:33:09] Not only you're

[00:33:10] investing in tech,

[00:33:10] then you're

[00:33:12] investing into

[00:33:13] productization,

[00:33:14] then you're

[00:33:14] investing into

[00:33:15] an ecosystem.

[00:33:15] Then you have

[00:33:16] to invest into

[00:33:16] your support

[00:33:17] network and

[00:33:17] post-sale support

[00:33:18] and all that.

[00:33:20] The payoff is

[00:33:22] longer and

[00:33:23] longer gestation

[00:33:24] period which

[00:33:24] the fund in

[00:33:25] itself may not

[00:33:26] allow.

[00:33:27] So you need a

[00:33:28] VC industry with

[00:33:29] layers of passing

[00:33:31] on and packaging

[00:33:31] and then moving

[00:33:32] on to the next

[00:33:34] because the LPs

[00:33:34] may not have

[00:33:35] the patience for

[00:33:37] that.

[00:33:38] But you should be

[00:33:38] able to package it

[00:33:39] and then give it to

[00:33:40] somebody else and

[00:33:40] then they should

[00:33:41] be able to take it

[00:33:41] up and they

[00:33:42] don't have a

[00:33:42] benchmark and

[00:33:43] milestones for

[00:33:44] them to understand

[00:33:45] how to judge

[00:33:47] this in these

[00:33:48] cases.

[00:33:49] That metrics are

[00:33:49] missing.

[00:33:50] So all of this

[00:33:51] has to be

[00:33:53] innovated from the

[00:33:53] front.

[00:33:54] That's what will

[00:33:54] take you.

[00:33:55] So these aspects

[00:33:56] actually come and

[00:33:58] we just think only

[00:33:58] from a solutions

[00:33:59] point of view,

[00:34:00] your product needs

[00:34:02] to inculcate all

[00:34:04] this, the

[00:34:07] metrics of the

[00:34:07] industry.

[00:34:08] Right.

[00:34:09] So yeah.

[00:34:10] Now we spoke

[00:34:10] about how you

[00:34:11] have to pick a

[00:34:12] solution, right?

[00:34:14] Or rather you

[00:34:14] have to deliver a

[00:34:15] solution and along

[00:34:16] the way you have

[00:34:16] to kind of

[00:34:17] product-based it

[00:34:18] and make it

[00:34:18] generally applicable

[00:34:19] for let's say

[00:34:20] the industry at

[00:34:21] large.

[00:34:21] For let's say

[00:34:23] someone who is

[00:34:23] at that phase,

[00:34:24] they've identified

[00:34:25] an interesting

[00:34:26] area to solve

[00:34:27] in.

[00:34:29] How do they

[00:34:30] source the right

[00:34:30] problem?

[00:34:31] How do they

[00:34:31] work with the

[00:34:32] right customers,

[00:34:32] the vendors?

[00:34:34] How do they go

[00:34:35] from that idea to

[00:34:36] product to company

[00:34:36] if you have some

[00:34:38] principles that you

[00:34:38] could share?

[00:34:39] Sure.

[00:34:40] I mean, when

[00:34:41] you're saying

[00:34:41] about idea, I

[00:34:43] still need a

[00:34:44] distinction whether

[00:34:44] it is a technical

[00:34:45] idea that they

[00:34:46] have or is it

[00:34:47] the business model

[00:34:48] idea that they

[00:34:49] have?

[00:34:49] Let's say like a

[00:34:49] broad technical

[00:34:50] idea, right?

[00:34:51] Someone says that

[00:34:51] they want to do

[00:34:52] something in

[00:34:52] computer vision,

[00:34:53] for example.

[00:34:54] Sure.

[00:34:54] Yeah.

[00:34:55] Perfect.

[00:34:55] Yeah.

[00:34:56] So one of the,

[00:34:57] again, I think I'll

[00:34:58] again pull back the

[00:34:59] point that I was

[00:35:00] saying many times

[00:35:02] really skilled

[00:35:03] engineers start

[00:35:05] thinking they

[00:35:06] could productize

[00:35:07] their skill.

[00:35:08] When you

[00:35:08] productize their

[00:35:09] skill, you

[00:35:09] could just

[00:35:09] become a

[00:35:10] service company,

[00:35:11] right?

[00:35:11] If you have that

[00:35:12] clarity, then it's

[00:35:13] great.

[00:35:13] If you want to

[00:35:14] just say that I'm

[00:35:14] going to build

[00:35:15] something and it's

[00:35:15] going to naturally

[00:35:16] going to become a

[00:35:16] product, that jump

[00:35:18] and, you know,

[00:35:19] that assumption,

[00:35:19] that visual thinking

[00:35:20] that we get into,

[00:35:21] like Elon Musk

[00:35:22] says, all of us

[00:35:24] had the fallacy,

[00:35:25] right?

[00:35:26] We, compared to my

[00:35:27] crowd around me, I

[00:35:28] have this edge and I

[00:35:30] see that somewhere

[00:35:30] this could be applied

[00:35:31] which is going to be a

[00:35:32] big problem.

[00:35:32] problem and then

[00:35:33] you have perceived

[00:35:34] that as a super

[00:35:35] large problem.

[00:35:36] When you yourself

[00:35:38] understand the

[00:35:39] economics of what

[00:35:40] it actually costs

[00:35:40] because you know,

[00:35:41] right, you yourself

[00:35:42] are a user in that

[00:35:43] case, right?

[00:35:43] So, which is not a

[00:35:45] luxury that you have

[00:35:46] when you are in a

[00:35:46] B2B, when you are

[00:35:47] moving to a B2B

[00:35:48] business, right?

[00:35:49] Some might originate

[00:35:51] from within the

[00:35:51] business, some might

[00:35:52] originate as a

[00:35:54] somebody who is just

[00:35:55] stepping fresh into

[00:35:56] the industry, right?

[00:35:58] I don't know how

[00:35:58] you will get those

[00:35:59] perspectives.

[00:36:00] You could work for

[00:36:01] an industry, you

[00:36:02] could work within

[00:36:02] that industry,

[00:36:04] whatnot, right?

[00:36:04] But you need to

[00:36:06] understand that you

[00:36:06] are not solving that

[00:36:07] technical problem for

[00:36:08] them, you are solving

[00:36:09] a business problem at

[00:36:10] the end of the day.

[00:36:10] It's a B2B business.

[00:36:11] You are talking to a

[00:36:12] system, not talking to

[00:36:13] an individual, right?

[00:36:16] If we don't make the

[00:36:17] distinction and then

[00:36:17] plan ourselves

[00:36:18] accordingly to that

[00:36:19] and then map out the

[00:36:21] nature of your

[00:36:21] problem, which is your

[00:36:22] customer, your

[00:36:23] customer is not that

[00:36:24] individual, your

[00:36:26] user might be an

[00:36:27] individual somewhere

[00:36:28] within that industry,

[00:36:29] but it has to pass

[00:36:30] through a whole

[00:36:31] system.

[00:36:32] The benefit, the

[00:36:33] sponsor, the

[00:36:34] again, someone who

[00:36:35] is going to monetize

[00:36:36] on top of that is

[00:36:36] all a system, right?

[00:36:38] So understanding

[00:36:40] how does it

[00:36:41] translate to a

[00:36:41] pain point to the

[00:36:42] system is where you

[00:36:43] need to start

[00:36:44] thinking about.

[00:36:45] And second, is the

[00:36:47] system aware of

[00:36:48] that problem?

[00:36:49] If you have to go

[00:36:50] teach the system

[00:36:51] about the problem,

[00:36:52] then that's a

[00:36:53] huge barrier you

[00:36:55] are looking at.

[00:36:56] They may not be

[00:36:57] solution aware, but

[00:36:58] they have to be

[00:36:58] problem aware.

[00:36:59] Problem aware.

[00:36:59] I'm talking about

[00:37:00] problem aware.

[00:37:01] Problem aware, see

[00:37:03] this also we face,

[00:37:04] right?

[00:37:04] So they are aware of

[00:37:05] the problem, but

[00:37:06] they believe that

[00:37:07] this is not

[00:37:08] solvable today, so

[00:37:09] they deprioritize it.

[00:37:11] You need to know

[00:37:11] why do they think

[00:37:12] it is not

[00:37:14] solvable today?

[00:37:14] I'm not saying, I

[00:37:15] mean, every

[00:37:15] disruption is breaking

[00:37:17] the thought for the

[00:37:18] customer, right?

[00:37:19] Otherwise, they are

[00:37:19] going with a person

[00:37:20] with a hammer always

[00:37:21] thinks with hammer,

[00:37:22] right?

[00:37:22] He will cut the

[00:37:23] tomato also with a

[00:37:24] hammer.

[00:37:25] We'll try to do that.

[00:37:26] That tendency is

[00:37:27] there for all of

[00:37:27] us.

[00:37:27] The industry also

[00:37:28] carries, of course,

[00:37:29] it's all human

[00:37:30] beings and they

[00:37:30] will carry that.

[00:37:31] The system also

[00:37:32] has that memory in

[00:37:33] itself.

[00:37:34] If you're breaking

[00:37:34] that, right, you

[00:37:36] need to know why

[00:37:36] is that perception?

[00:37:38] What did he try

[00:37:39] before?

[00:37:40] Why did he fail in

[00:37:40] those cases?

[00:37:41] Why were they so

[00:37:43] negative about, for

[00:37:44] us, it was vision.

[00:37:45] The moment you say

[00:37:46] with vision, they

[00:37:46] will be like, no,

[00:37:47] I know.

[00:37:48] We have tried this

[00:37:49] morning to evening,

[00:37:49] lighting changes, it

[00:37:50] doesn't work, right?

[00:37:52] It's not eliminating

[00:37:53] a person for me.

[00:37:53] I have to put a

[00:37:55] person there and

[00:37:56] at least if it's

[00:37:56] periodic, it's fine.

[00:37:58] For every 10

[00:37:59] minutes one, for

[00:37:59] sure, I need a

[00:38:00] person to come in

[00:38:01] there.

[00:38:01] But if I do it

[00:38:02] for 10 systems,

[00:38:02] I'm replacing 10

[00:38:03] people with 3

[00:38:04] people, possibly.

[00:38:05] I'm just giving an

[00:38:06] example, right?

[00:38:07] Then at least 7

[00:38:07] people reduction is

[00:38:08] there.

[00:38:09] So, okay, fine.

[00:38:10] That's a system that

[00:38:11] I can adopt.

[00:38:11] But you don't know

[00:38:12] whether it is 1

[00:38:12] minute once you need

[00:38:13] it or 10 minute

[00:38:14] once.

[00:38:14] So, you need to have

[00:38:15] the person

[00:38:15] continuously there.

[00:38:18] So, erratic

[00:38:19] need for availability

[00:38:20] of people, if you

[00:38:21] still have, your

[00:38:22] technology still

[00:38:22] demands that.

[00:38:23] That's not going

[00:38:24] to work out.

[00:38:24] I'm just giving

[00:38:24] one insight like

[00:38:25] that.

[00:38:26] There will be

[00:38:26] several other

[00:38:27] things, right?

[00:38:27] How does it roll

[00:38:28] down to his

[00:38:29] supply chain?

[00:38:30] How does it roll

[00:38:31] down to his

[00:38:31] sales cycle?

[00:38:32] What is the

[00:38:33] benefit that he's

[00:38:33] going to get by

[00:38:34] doing this?

[00:38:35] How is he going

[00:38:35] to push?

[00:38:36] Where are the

[00:38:37] organization goals

[00:38:37] set?

[00:38:38] How is the budget

[00:38:38] system working

[00:38:40] within the

[00:38:41] organization?

[00:38:42] And sometimes

[00:38:43] as bad as

[00:38:44] the person who

[00:38:45] was sponsoring

[00:38:46] this inside the

[00:38:47] organization moves

[00:38:48] out.

[00:38:49] right?

[00:38:50] You're gone.

[00:38:52] Things like that

[00:38:53] just, you know.

[00:38:55] So, then how do

[00:38:56] you insulate

[00:38:57] yourself from

[00:38:58] those risks?

[00:38:59] How do you

[00:38:59] parallelize to

[00:39:00] have multiple?

[00:39:01] And one

[00:39:03] customer here,

[00:39:04] again for our

[00:39:05] industry I'm saying

[00:39:05] in a B2B,

[00:39:07] one customer

[00:39:09] is multiple

[00:39:10] accounts or

[00:39:11] one account is

[00:39:11] multiple customers.

[00:39:12] One account is

[00:39:13] multiple customers,

[00:39:13] right?

[00:39:14] You might have a

[00:39:14] label of, let's

[00:39:16] say, somebody

[00:39:16] like an

[00:39:17] automotive giant

[00:39:18] like GM or

[00:39:18] Toyota or

[00:39:19] Ford or

[00:39:20] somebody.

[00:39:21] Within them,

[00:39:21] there will be

[00:39:21] multiple different

[00:39:22] customer ranges.

[00:39:24] And so many

[00:39:24] departments,

[00:39:25] divisions and

[00:39:27] across them,

[00:39:28] right?

[00:39:28] And again, you

[00:39:29] need to know

[00:39:29] what layer you

[00:39:30] are selling to.

[00:39:32] If you're

[00:39:32] somebody who's

[00:39:33] selling a wire

[00:39:33] and a cable,

[00:39:34] you're selling at

[00:39:34] the plant

[00:39:35] engineer level.

[00:39:36] If you're

[00:39:37] talking about

[00:39:37] a process and

[00:39:38] a transformation,

[00:39:39] you're talking

[00:39:39] to somebody

[00:39:40] else.

[00:39:41] If you're

[00:39:41] talking about

[00:39:42] business model

[00:39:43] modification,

[00:39:43] then you're

[00:39:43] talking to

[00:39:44] the CFO.

[00:39:45] So you

[00:39:45] have different

[00:39:45] customers within

[00:39:46] your, whose

[00:39:48] problem are

[00:39:48] you solving?

[00:39:49] It could be

[00:39:50] technology, it

[00:39:51] could be the

[00:39:51] same vision,

[00:39:52] right?

[00:39:53] Is it a QA

[00:39:54] problem that

[00:39:54] you're solving?

[00:39:56] You need to

[00:39:56] understand

[00:39:57] procurement,

[00:39:57] basically.

[00:39:58] You need to

[00:39:58] understand structure

[00:39:59] of the business.

[00:40:00] Right.

[00:40:00] Because it's a

[00:40:01] well-considered

[00:40:03] decision to use

[00:40:04] one thing or

[00:40:05] another.

[00:40:06] It's not like an

[00:40:07] impulse purchase.

[00:40:08] It's not an

[00:40:08] impulse purchase.

[00:40:09] Right?

[00:40:09] So you need to

[00:40:10] understand what

[00:40:11] are the, you

[00:40:12] know, the

[00:40:12] complete hub and

[00:40:13] scope of all

[00:40:14] of this stuff.

[00:40:15] Correct.

[00:40:15] And who you're

[00:40:16] selling to.

[00:40:17] Correct.

[00:40:17] And you also

[00:40:18] need to

[00:40:18] understand how

[00:40:19] much is their

[00:40:20] experimental budget.

[00:40:22] That's easier

[00:40:23] said than done.

[00:40:25] It's harder,

[00:40:26] they will not

[00:40:26] reveal, but you

[00:40:28] need to have,

[00:40:28] so sometimes

[00:40:29] having that

[00:40:29] inside leg

[00:40:31] allows, right?

[00:40:32] And you'll

[00:40:34] also have a

[00:40:34] problem of how

[00:40:35] would you sell

[00:40:36] your tech to

[00:40:37] the, you need

[00:40:38] the right ears,

[00:40:39] right person,

[00:40:40] you need to go

[00:40:41] right?

[00:40:41] You can't

[00:40:41] just interface

[00:40:42] with the

[00:40:42] purchase guy

[00:40:42] and then hope

[00:40:43] that this will

[00:40:44] work inwards.

[00:40:45] Who's the

[00:40:45] end customer

[00:40:46] within the

[00:40:47] organization,

[00:40:48] right?

[00:40:49] And who

[00:40:49] realizes this

[00:40:50] very clearly?

[00:40:51] Right.

[00:40:51] So this

[00:40:52] comes back

[00:40:52] to the,

[00:40:53] again,

[00:40:53] the problem

[00:40:53] of have

[00:40:54] they realized

[00:40:54] that,

[00:40:55] do they realize

[00:40:56] that it is a

[00:40:57] soluble problem

[00:40:58] or do they

[00:40:58] realize that it

[00:40:58] is an unsoluble

[00:40:59] problem?

[00:41:00] If they think

[00:41:00] that it's a

[00:41:00] soluble problem,

[00:41:01] you don't

[00:41:01] have much scope.

[00:41:02] Most probably

[00:41:03] they're popularized

[00:41:03] it, you'll be

[00:41:04] one among the

[00:41:04] competition.

[00:41:05] If you're okay,

[00:41:06] so even there,

[00:41:07] are you trying

[00:41:08] to compete

[00:41:08] with someone

[00:41:09] else because

[00:41:09] you have a

[00:41:09] better tech

[00:41:11] or are you

[00:41:12] competing with

[00:41:14] an unsolved

[00:41:15] problem?

[00:41:16] So that also

[00:41:17] you should be

[00:41:18] very clear

[00:41:18] about,

[00:41:18] right?

[00:41:19] So we

[00:41:19] looked at,

[00:41:20] for us,

[00:41:21] we looked at

[00:41:21] that it's a

[00:41:22] fundamental

[00:41:22] technology.

[00:41:23] It took a

[00:41:23] lot more

[00:41:23] time than

[00:41:24] what we

[00:41:26] anticipated

[00:41:26] or kind of

[00:41:27] we were

[00:41:27] prepared for

[00:41:28] it,

[00:41:28] but what the

[00:41:30] stakeholders

[00:41:30] were expecting

[00:41:31] out of

[00:41:31] us.

[00:41:33] But for

[00:41:34] us,

[00:41:36] it's a

[00:41:36] fundamental

[00:41:36] technology

[00:41:37] which means

[00:41:37] we can't

[00:41:38] be talking

[00:41:38] to the

[00:41:38] system.

[00:41:39] That is

[00:41:40] meant for

[00:41:40] procuring

[00:41:42] competitive

[00:41:42] technologies

[00:41:43] that are already

[00:41:44] available.

[00:41:45] You need to

[00:41:45] talk for

[00:41:45] somebody who

[00:41:46] is there

[00:41:46] within the

[00:41:46] organization

[00:41:47] who is

[00:41:47] thinking of

[00:41:48] future

[00:41:48] transformations,

[00:41:49] who is

[00:41:49] thinking

[00:41:49] five years

[00:41:50] ahead.

[00:41:50] So that

[00:41:51] goes to

[00:41:51] the leadership

[00:41:51] all the

[00:41:52] way up.

[00:41:53] It has

[00:41:53] to trickle

[00:41:54] top-down.

[00:41:55] Is it

[00:41:55] bottom-up

[00:41:56] or a

[00:41:56] top-down

[00:41:56] problem?

[00:41:57] Some things,

[00:41:58] if you go

[00:41:58] top-down,

[00:41:59] it will not

[00:41:59] work.

[00:42:00] You need

[00:42:01] to go

[00:42:01] bottom-up.

[00:42:02] Something

[00:42:02] you have

[00:42:03] to hit

[00:42:03] both.

[00:42:03] So you

[00:42:04] need to

[00:42:04] know where

[00:42:04] do you

[00:42:05] fall.

[00:42:05] The problem

[00:42:06] that you're

[00:42:06] taking,

[00:42:06] where do

[00:42:06] you fall?

[00:42:08] So that

[00:42:09] systemic

[00:42:10] understanding

[00:42:10] of the

[00:42:10] customer and

[00:42:11] the industry

[00:42:12] space that

[00:42:12] you're

[00:42:12] dealing with.

[00:42:13] And by

[00:42:13] the way,

[00:42:14] each industry

[00:42:14] will behave

[00:42:14] differently.

[00:42:15] I can say

[00:42:16] manufacturing,

[00:42:16] but the

[00:42:16] manufacturing of

[00:42:17] food is

[00:42:17] very different

[00:42:18] from

[00:42:18] manufacturing

[00:42:18] for

[00:42:19] automotive,

[00:42:20] automotive

[00:42:20] tire one,

[00:42:21] OEM,

[00:42:22] versus your

[00:42:23] electronics.

[00:42:25] Each of

[00:42:25] them have a

[00:42:25] different process,

[00:42:26] different setups,

[00:42:27] different layers

[00:42:27] of decision

[00:42:28] making.

[00:42:28] Right.

[00:42:28] So does it

[00:42:29] make sense

[00:42:30] to first

[00:42:30] narrow down

[00:42:31] what field

[00:42:32] I want to

[00:42:32] solve for

[00:42:34] and then

[00:42:34] keep them

[00:42:34] parallel.

[00:42:36] You can't

[00:42:36] leave out

[00:42:37] your technical

[00:42:37] advantage.

[00:42:38] Because that

[00:42:39] itself consumes

[00:42:39] a lot of

[00:42:40] effort and

[00:42:40] time and

[00:42:41] understanding

[00:42:42] the foundations

[00:42:42] and fundamentals

[00:42:43] of what

[00:42:43] you're

[00:42:43] actually

[00:42:43] building.

[00:42:44] While

[00:42:45] that is

[00:42:45] at it,

[00:42:46] don't

[00:42:46] sequentialize

[00:42:47] this.

[00:42:47] Keep

[00:42:47] this

[00:42:48] parallel.

[00:42:48] You need

[00:42:49] to also

[00:42:49] go understand

[00:42:50] the problem.

[00:42:51] As much

[00:42:52] as you're

[00:42:53] excited by

[00:42:53] understanding

[00:42:54] the tech

[00:42:55] and your

[00:42:55] ability to

[00:42:56] apply the

[00:42:56] tech,

[00:42:57] you don't

[00:42:58] put that

[00:42:58] much energy

[00:42:58] on knowing

[00:42:59] the problem.

[00:43:00] Spend that

[00:43:01] as part of

[00:43:01] the problem

[00:43:02] and understand,

[00:43:03] think that

[00:43:03] this whole

[00:43:04] structure is

[00:43:04] the problem.

[00:43:05] Right.

[00:43:05] This is a

[00:43:06] typical engineer's

[00:43:06] mindset.

[00:43:07] This is a

[00:43:07] typical engineer's

[00:43:08] mindset.

[00:43:09] Right.

[00:43:09] Which I'm

[00:43:09] also at

[00:43:11] blame for.

[00:43:13] one follow-up

[00:43:14] on this and

[00:43:14] then we'll

[00:43:15] move on to

[00:43:15] some of

[00:43:15] the technical

[00:43:16] challenges

[00:43:16] that you've

[00:43:17] solved at

[00:43:17] SineDoc.

[00:43:19] How do

[00:43:19] you find

[00:43:20] the most

[00:43:20] profitable

[00:43:21] problems to

[00:43:21] solve?

[00:43:24] Interesting.

[00:43:25] How do I

[00:43:26] find the

[00:43:26] most profitable

[00:43:27] problem to

[00:43:27] solve?

[00:43:29] Let's say

[00:43:29] you have

[00:43:30] four,

[00:43:30] five,

[00:43:31] seven years

[00:43:31] in the

[00:43:32] industry.

[00:43:32] You have

[00:43:33] a certain

[00:43:33] skill set.

[00:43:34] Now you

[00:43:35] want to be

[00:43:35] able to

[00:43:36] map the

[00:43:37] skill set

[00:43:37] to a

[00:43:38] problem

[00:43:38] that the

[00:43:38] industry

[00:43:39] is paying

[00:43:39] top dollar

[00:43:39] for.

[00:43:44] I'm not

[00:43:45] saying that's

[00:43:45] wrong or

[00:43:46] right.

[00:43:48] I'm not

[00:43:48] so good at

[00:43:49] competing.

[00:43:50] It's a

[00:43:51] distraction for

[00:43:52] me.

[00:43:55] And I

[00:43:56] don't go

[00:43:56] with,

[00:43:57] so anything

[00:43:57] that peaks

[00:43:58] in its,

[00:43:59] this is a

[00:44:00] challenge,

[00:44:01] right?

[00:44:01] Anything

[00:44:02] peaks at

[00:44:03] its

[00:44:03] remuneration,

[00:44:03] has had an

[00:44:04] exponential

[00:44:04] growth,

[00:44:05] will also

[00:44:06] have a

[00:44:06] dip.

[00:44:06] Because that's

[00:44:07] the first

[00:44:07] thing the

[00:44:07] industry would

[00:44:08] want to

[00:44:08] replace.

[00:44:10] Right?

[00:44:11] Volume is

[00:44:12] very different

[00:44:12] from

[00:44:13] profitability.

[00:44:15] which means

[00:44:16] it's expensive

[00:44:16] for someone.

[00:44:17] That's a

[00:44:18] profit that

[00:44:18] the customer

[00:44:18] can eat.

[00:44:20] Right?

[00:44:21] You have to

[00:44:22] borrow that

[00:44:22] for you.

[00:44:23] If you have

[00:44:24] margins exceptionally

[00:44:25] very high,

[00:44:26] it either

[00:44:27] evokes more

[00:44:27] competition within

[00:44:28] your space,

[00:44:29] which means

[00:44:29] you are

[00:44:30] hitting the

[00:44:30] saturation

[00:44:31] point.

[00:44:31] It's going

[00:44:32] to start,

[00:44:33] henceforth,

[00:44:34] it's going to

[00:44:34] get hyper

[00:44:34] commoditized.

[00:44:35] That's why

[00:44:36] you're getting

[00:44:36] into, so

[00:44:36] the business

[00:44:37] will have

[00:44:38] a transformation.

[00:44:39] You're at

[00:44:39] the edge of

[00:44:40] the business

[00:44:40] transformation.

[00:44:41] Right?

[00:44:42] Like,

[00:44:42] starting

[00:44:43] Swiggy

[00:44:44] or Uber

[00:44:45] in 2000s,

[00:44:46] early 2000s

[00:44:47] is very

[00:44:48] different from

[00:44:48] starting in

[00:44:50] 2011s and

[00:44:51] 12s and

[00:44:51] 13s and

[00:44:52] so on.

[00:44:53] The

[00:44:53] business is,

[00:44:54] difficulties

[00:44:55] of the

[00:44:55] business is

[00:44:55] very different.

[00:44:56] The journey

[00:44:57] that Flipkart

[00:44:58] went through

[00:44:58] in India

[00:44:59] or Amazon

[00:45:00] went through

[00:45:00] globally.

[00:45:02] We always

[00:45:03] talk about

[00:45:03] the first

[00:45:03] mover

[00:45:04] advantage

[00:45:04] and there

[00:45:05] is a lot

[00:45:06] more

[00:45:06] first mover

[00:45:06] disadvantage

[00:45:07] because you

[00:45:07] build a

[00:45:08] whole ecosystem

[00:45:08] around you.

[00:45:09] That's never

[00:45:09] talked about

[00:45:10] and most

[00:45:11] of the

[00:45:11] products that

[00:45:11] you use

[00:45:11] are not

[00:45:12] built by

[00:45:12] the first

[00:45:13] guys,

[00:45:14] including

[00:45:14] your Google

[00:45:14] and so

[00:45:15] on and

[00:45:15] so forth.

[00:45:18] So profitability

[00:45:18] is something

[00:45:19] that you look

[00:45:20] in terms of

[00:45:21] size of the

[00:45:22] market that you

[00:45:22] can capture

[00:45:22] than, I

[00:45:24] look at it

[00:45:24] from that

[00:45:25] point of view,

[00:45:26] than the depth

[00:45:26] of margin

[00:45:27] that you can

[00:45:27] get.

[00:45:27] So you can

[00:45:28] think about,

[00:45:29] I will sell

[00:45:29] only 100 pieces

[00:45:30] every year,

[00:45:30] but I can

[00:45:30] have a large

[00:45:31] depth of

[00:45:32] profit.

[00:45:32] That's a

[00:45:33] car.niche

[00:45:34] area where

[00:45:37] they are

[00:45:37] willing to

[00:45:38] do premium.

[00:45:40] So if I

[00:45:40] work backwards

[00:45:41] from what

[00:45:42] is that you

[00:45:43] could replace

[00:45:43] at the same

[00:45:44] cost as

[00:45:44] what the

[00:45:45] customer has

[00:45:45] today,

[00:45:48] of which

[00:45:49] you could

[00:45:49] make maximum

[00:45:49] profits.

[00:45:53] So that's

[00:45:54] something that

[00:45:55] we should

[00:45:57] trace back

[00:45:58] in our

[00:45:59] customer

[00:46:01] journey.

[00:46:01] So if I

[00:46:02] start looking

[00:46:02] at, is

[00:46:03] this the

[00:46:04] product that

[00:46:04] I can

[00:46:05] maximum

[00:46:08] earn, where

[00:46:09] the margins

[00:46:10] are very

[00:46:10] high, like

[00:46:11] if I am

[00:46:12] replacing a

[00:46:13] human being,

[00:46:13] what kind of

[00:46:14] human being

[00:46:14] am I

[00:46:14] replacing?

[00:46:15] What kind

[00:46:16] of value

[00:46:17] am I

[00:46:17] replacing?

[00:46:17] What kind

[00:46:18] of labor

[00:46:18] am I

[00:46:18] replacing?

[00:46:19] What's

[00:46:19] their

[00:46:19] charges?

[00:46:20] Is it

[00:46:20] specialized

[00:46:20] work?

[00:46:21] Like for

[00:46:21] example,

[00:46:22] there are

[00:46:23] tasks that

[00:46:23] need eight

[00:46:24] shifts versus

[00:46:25] three shifts

[00:46:27] where it's

[00:46:28] ergonomically

[00:46:28] a huge

[00:46:29] issue.

[00:46:30] They can't

[00:46:30] underbelly of

[00:46:31] a car,

[00:46:31] you can't

[00:46:31] just keep

[00:46:33] doing all

[00:46:34] through the

[00:46:34] day.

[00:46:34] So maximum

[00:46:35] two hours.

[00:46:36] After that,

[00:46:37] because it's

[00:46:37] a continuously

[00:46:38] moving car and

[00:46:39] then you're

[00:46:39] continuously

[00:46:40] operating on

[00:46:40] it, you

[00:46:41] have to keep

[00:46:41] moving like

[00:46:42] a honeybee.

[00:46:43] You're

[00:46:44] continuously

[00:46:44] on it.

[00:46:45] So after

[00:46:46] two hours,

[00:46:46] you can't

[00:46:47] sustain

[00:46:47] doing that.

[00:46:48] Then another

[00:46:49] person comes.

[00:46:49] So the cost

[00:46:50] for doing that

[00:46:51] activity is very

[00:46:51] high for the

[00:46:53] customer.

[00:46:53] That's something

[00:46:54] that you borrow

[00:46:54] and then automate

[00:46:55] first.

[00:46:55] That's a

[00:46:55] luxury.

[00:46:56] Second

[00:46:56] thing, I

[00:46:57] also look

[00:46:57] at initial

[00:46:58] days you

[00:46:59] might want

[00:47:00] to have,

[00:47:00] you will

[00:47:01] not be able

[00:47:01] to make

[00:47:02] your margins

[00:47:02] in spite of

[00:47:03] you being

[00:47:03] exponentially

[00:47:04] expensive.

[00:47:06] Why so?

[00:47:07] Your supply

[00:47:08] chain itself

[00:47:08] is not

[00:47:10] mature.

[00:47:10] Your volumes

[00:47:11] are not

[00:47:11] mature.

[00:47:12] I will get

[00:47:12] at a premium

[00:47:13] compared to

[00:47:13] someone else,

[00:47:14] some other

[00:47:14] guy with the

[00:47:15] same components,

[00:47:16] some other guy

[00:47:17] who will actually

[00:47:17] give them a

[00:47:17] volume business.

[00:47:19] So my net

[00:47:20] cost of the

[00:47:21] solution is very

[00:47:22] high.

[00:47:22] Yet the customer

[00:47:23] is supposed to be

[00:47:24] profitable on and

[00:47:24] above that.

[00:47:25] So you

[00:47:26] need to look

[00:47:26] at the

[00:47:27] roadsters

[00:47:27] of your

[00:47:30] luxury thing.

[00:47:32] Often

[00:47:33] initial picks

[00:47:34] like, and

[00:47:34] you need a

[00:47:35] patient customer

[00:47:36] who is okay

[00:47:36] with this

[00:47:37] failing.

[00:47:38] Often that

[00:47:38] will be

[00:47:38] somebody who

[00:47:39] is trying to

[00:47:39] build an

[00:47:40] image through

[00:47:40] your product.

[00:47:42] and it's a

[00:47:42] very high

[00:47:43] technology product.

[00:47:44] So they want

[00:47:44] to showcase

[00:47:45] this and

[00:47:46] they want to

[00:47:46] add the

[00:47:47] value of

[00:47:48] their branding

[00:47:48] more.

[00:47:49] So that's

[00:47:50] another way

[00:47:51] you could

[00:47:51] tap initially.

[00:47:52] So otherwise

[00:47:54] I am not

[00:47:56] very clear

[00:47:57] about where

[00:47:59] will you get

[00:47:59] your high

[00:48:00] margins and

[00:48:01] high profitability

[00:48:02] systems within

[00:48:03] that.

[00:48:03] That's more

[00:48:03] of your

[00:48:03] business model

[00:48:04] working internally.

[00:48:06] So how we

[00:48:06] started with

[00:48:07] is we

[00:48:08] fixated the

[00:48:08] price point.

[00:48:09] This is what

[00:48:10] the customer

[00:48:10] is willing to

[00:48:11] pay.

[00:48:11] This is what

[00:48:11] they can

[00:48:12] afford.

[00:48:12] And then we

[00:48:13] worked backwards

[00:48:14] from there.

[00:48:14] And that was a

[00:48:15] conservative amount

[00:48:15] by the time

[00:48:16] we caught up

[00:48:17] the bare

[00:48:17] minimum

[00:48:18] has actually

[00:48:21] grown higher

[00:48:22] in the value

[00:48:22] because of

[00:48:23] COVID and

[00:48:24] post-COVID

[00:48:24] and supply

[00:48:24] chain issues

[00:48:25] and so on

[00:48:26] so forth.

[00:48:26] So sometimes

[00:48:28] the serendipity

[00:48:28] also works

[00:48:29] but otherwise

[00:48:29] you just look

[00:48:30] at how much

[00:48:30] are you

[00:48:31] consolidating

[00:48:33] as expenses

[00:48:33] and then you

[00:48:34] could actually

[00:48:34] take that away.

[00:48:36] The second

[00:48:37] way to think

[00:48:38] is often

[00:48:41] if you can

[00:48:42] get into

[00:48:42] services added

[00:48:43] on top of

[00:48:43] your hardware

[00:48:44] platform

[00:48:46] that's a great

[00:48:46] value addition

[00:48:47] that you can

[00:48:48] start looking

[00:48:49] at.

[00:48:50] So that's a

[00:48:50] good margin

[00:48:51] business for

[00:48:52] you.

[00:48:52] Profitable

[00:48:52] business for

[00:48:53] you.

[00:48:53] So I

[00:48:54] wouldn't think

[00:48:54] about hitting

[00:48:55] the price point

[00:48:56] right is a

[00:48:57] big problem

[00:48:58] right.

[00:48:58] So can we

[00:48:59] get the

[00:48:59] price point

[00:49:00] right?

[00:49:00] Can we

[00:49:00] get the

[00:49:00] price point

[00:49:01] right?

[00:49:01] What if I

[00:49:02] have under

[00:49:02] priced or

[00:49:03] what if I

[00:49:03] have over

[00:49:03] priced?

[00:49:04] There is a

[00:49:04] bare minimum

[00:49:04] price that

[00:49:05] you have to

[00:49:05] work as a

[00:49:05] model.

[00:49:06] And I

[00:49:06] start looking

[00:49:07] at let

[00:49:07] your customer

[00:49:08] set the

[00:49:08] upper price.

[00:49:10] Keep your

[00:49:11] services open

[00:49:12] so that you

[00:49:13] can extend

[00:49:13] and then later

[00:49:14] you can just

[00:49:14] keep adding

[00:49:14] that more

[00:49:15] and more

[00:49:15] and more.

[00:49:16] So when I

[00:49:17] said the

[00:49:17] most profitable

[00:49:18] ideas I

[00:49:19] didn't mean

[00:49:19] it in the

[00:49:19] precise sense

[00:49:20] of the word

[00:49:21] but this

[00:49:22] was an

[00:49:22] education for

[00:49:23] sure.

[00:49:23] right.

[00:49:25] Yeah I

[00:49:25] mean since

[00:49:26] you talk

[00:49:26] about price

[00:49:27] right I

[00:49:27] mean there's

[00:49:27] technical

[00:49:28] innovation and

[00:49:29] then manufacturing

[00:49:29] is a whole

[00:49:30] other thing

[00:49:30] right.

[00:49:32] Elon Musk

[00:49:33] himself mentioned

[00:49:34] this that you

[00:49:34] know manufacturing

[00:49:35] is a whole

[00:49:36] different beast

[00:49:38] altogether right.

[00:49:39] So how does

[00:49:41] one understand

[00:49:42] the nuances

[00:49:43] of manufacturing

[00:49:44] enough to

[00:49:45] take product

[00:49:46] to market

[00:49:46] right.

[00:49:47] I mean

[00:49:47] whether it is

[00:49:48] interfacing with

[00:49:49] vendors whether

[00:49:50] it is like

[00:49:50] picking what

[00:49:51] materials you're

[00:49:52] going to use

[00:49:54] yeah the

[00:49:54] whole lot

[00:49:55] of it.

[00:49:56] Yeah.

[00:49:58] We could spend

[00:49:58] a couple of

[00:49:59] hours on that

[00:49:59] but you know

[00:50:00] if you could

[00:50:01] just like

[00:50:02] cover the

[00:50:03] broad contours

[00:50:04] of it.

[00:50:04] So

[00:50:05] manufacturability

[00:50:09] I think

[00:50:09] again one

[00:50:11] fundamental

[00:50:12] question that

[00:50:12] you have to

[00:50:13] ask because

[00:50:14] it depends

[00:50:15] on where is

[00:50:15] the background

[00:50:16] that you're

[00:50:16] coming from

[00:50:16] right.

[00:50:16] So if

[00:50:17] person who

[00:50:17] is within

[00:50:18] the industry

[00:50:18] even when

[00:50:19] I was within

[00:50:19] the industry

[00:50:20] I didn't

[00:50:20] understand

[00:50:20] manufacturability

[00:50:21] to this

[00:50:21] extent that I

[00:50:22] understand today

[00:50:22] that's something

[00:50:23] you have to

[00:50:23] bite a little

[00:50:24] bit.

[00:50:24] Right.

[00:50:25] You don't

[00:50:25] understand the

[00:50:26] components and

[00:50:26] so on.

[00:50:27] Because that's

[00:50:27] usually even

[00:50:28] within those

[00:50:28] industries these

[00:50:29] are like marked

[00:50:31] different roles

[00:50:32] that are kept

[00:50:32] and there's a

[00:50:33] whole team

[00:50:33] behind it and

[00:50:34] then comes

[00:50:35] around that

[00:50:35] right.

[00:50:35] So but the

[00:50:36] moment you start

[00:50:37] understanding the

[00:50:37] pain of

[00:50:38] transactions and

[00:50:40] how hard it

[00:50:41] is to you just

[00:50:42] think that oh I

[00:50:42] pay something that

[00:50:43] I'm going to

[00:50:43] get you won't

[00:50:45] get it.

[00:50:46] Right.

[00:50:46] Like one of the

[00:50:47] first experience

[00:50:48] that I had is

[00:50:49] the first

[00:50:50] robotic arm that

[00:50:51] we bought from

[00:50:52] one of the

[00:50:52] vendors from

[00:50:52] within the

[00:50:53] country very

[00:50:55] initial in

[00:50:55] 2019 around

[00:50:57] after we

[00:50:58] raised our

[00:50:58] funding in

[00:50:58] August the

[00:51:00] second of

[00:51:00] the year

[00:51:01] around October

[00:51:02] we've you know

[00:51:03] placed the order

[00:51:04] the lead time is

[00:51:05] at least you

[00:51:06] know around 90

[00:51:07] days or so

[00:51:08] and that person

[00:51:09] got the order

[00:51:11] from us and

[00:51:12] then he left

[00:51:13] the company

[00:51:13] right left

[00:51:14] that the

[00:51:15] the vendors

[00:51:16] company and

[00:51:17] he just shipped

[00:51:18] the wrong

[00:51:18] product to

[00:51:18] us it's a

[00:51:19] wrong sale.

[00:51:21] Wow.

[00:51:21] Right.

[00:51:22] And that time

[00:51:23] it's such a

[00:51:23] small amount

[00:51:24] right that 770

[00:51:24] K and one

[00:51:26] robot setup

[00:51:26] in itself costs

[00:51:27] you around

[00:51:27] 300 K a lot

[00:51:28] of times by

[00:51:28] the way this

[00:51:29] is one more

[00:51:29] point that I

[00:51:30] had to kind

[00:51:30] of highlight

[00:51:31] many times

[00:51:32] you would want

[00:51:32] to when you're

[00:51:33] dealing with

[00:51:34] this complicated

[00:51:35] industries you

[00:51:35] want to start

[00:51:36] with your own

[00:51:36] money try to

[00:51:38] do that and

[00:51:38] so and so

[00:51:41] it's possible

[00:51:42] in your software

[00:51:43] because your

[00:51:43] investment is just

[00:51:44] your laptop and

[00:51:45] then you could

[00:51:46] people and

[00:51:47] laptop and

[00:51:47] people could be

[00:51:48] your co-founders

[00:51:49] or it could be

[00:51:49] your founding

[00:51:50] member teams

[00:51:51] and even

[00:51:52] yourself for that

[00:51:53] case right

[00:51:54] that's not the

[00:51:54] case when you

[00:51:55] come to a

[00:51:56] proper industrial

[00:51:56] gadget and

[00:51:57] things and

[00:51:58] your trials

[00:51:59] that you show

[00:51:59] their customer

[00:52:00] will not even

[00:52:00] look at it

[00:52:01] they need a

[00:52:02] system that

[00:52:02] can work for

[00:52:03] next 5 years

[00:52:04] so you need to

[00:52:05] begin with

[00:52:05] something that

[00:52:06] you can

[00:52:06] can off the

[00:52:07] shelf which

[00:52:07] is proven for

[00:52:09] working for

[00:52:09] 5 years and

[00:52:10] so on and so

[00:52:10] forth but

[00:52:12] when you have to

[00:52:12] deal with all

[00:52:13] these complexities

[00:52:14] how you could

[00:52:15] work with those

[00:52:17] variations that

[00:52:18] can happen

[00:52:18] then you will

[00:52:19] know how much

[00:52:20] of effort that

[00:52:20] is needed

[00:52:20] how the

[00:52:21] thinking should

[00:52:21] be what we

[00:52:22] should do

[00:52:22] around right

[00:52:25] now you will

[00:52:26] see that there

[00:52:26] is time as a

[00:52:27] factor that comes

[00:52:28] in so just

[00:52:28] because you pay

[00:52:29] the money you're

[00:52:30] not going to get

[00:52:30] that what are

[00:52:31] you paying the

[00:52:31] money for how

[00:52:33] much cost

[00:52:33] how much of

[00:52:34] premium you're

[00:52:34] paying for each

[00:52:35] of this hardware

[00:52:36] and you also need

[00:52:38] to understand

[00:52:38] that those guys

[00:52:41] this is on the

[00:52:41] supply channel

[00:52:42] not I'm talking

[00:52:42] about not it into

[00:52:46] when you are

[00:52:47] paying this

[00:52:48] premium you

[00:52:49] always remember

[00:52:50] to think that

[00:52:51] he's paying the

[00:52:52] premium where is

[00:52:53] this premium

[00:52:53] coming from is

[00:52:54] he trying to

[00:52:54] earn most out of

[00:52:55] you you're a

[00:52:56] startup they

[00:52:56] should have

[00:52:56] given you a

[00:52:56] discount all

[00:52:57] this you can

[00:52:57] think about but

[00:52:58] for him it

[00:52:58] costs more

[00:53:00] because a TVS

[00:53:01] who might buy

[00:53:01] 50 robots

[00:53:02] versus you who

[00:53:03] is buying one

[00:53:04] robot both of

[00:53:05] you would need

[00:53:05] one application

[00:53:06] engineer full time

[00:53:07] being kept

[00:53:07] there but the

[00:53:08] cost of the

[00:53:09] application

[00:53:09] engineer is

[00:53:10] shared between

[00:53:10] 50 robots

[00:53:11] for you it

[00:53:12] is shared

[00:53:12] between one

[00:53:13] robot even

[00:53:14] if he doesn't

[00:53:14] get a single

[00:53:15] robot sale for

[00:53:16] next six months

[00:53:17] he's not going

[00:53:18] to accommodate

[00:53:18] you even though

[00:53:19] you're giving

[00:53:20] you know you're

[00:53:21] saying that I'm

[00:53:21] going to give

[00:53:21] him one one

[00:53:22] robot possibly

[00:53:22] every three

[00:53:23] months once

[00:53:23] he's not going

[00:53:24] to accommodate

[00:53:24] you you'll rather

[00:53:25] take that sales

[00:53:27] time and effort

[00:53:28] and then apply for

[00:53:29] a customer who

[00:53:30] would have a

[00:53:31] potential

[00:53:32] TVS may not

[00:53:33] buy a 50 robots

[00:53:33] up front but he

[00:53:34] would buy one

[00:53:34] robot but the

[00:53:35] one robot that

[00:53:36] there is an

[00:53:36] assured potential

[00:53:37] of 50

[00:53:38] translating to

[00:53:39] 50 that's

[00:53:40] going to happen

[00:53:40] for you

[00:53:42] right you need

[00:53:43] to think of

[00:53:43] vendors as your

[00:53:44] investors right

[00:53:47] every stakeholder

[00:53:48] essentially becomes

[00:53:48] an investor from

[00:53:49] the front right

[00:53:51] somewhere somebody

[00:53:52] should be willing

[00:53:52] to invest that

[00:53:53] A's extra cost

[00:53:54] he has budgeted

[00:53:55] in his margin

[00:53:56] saying that I'm

[00:53:56] going to give

[00:53:56] only these many

[00:53:57] hours of

[00:53:57] application engineering

[00:53:58] or post sales

[00:53:59] services or sales

[00:54:00] effort for you

[00:54:01] but when I am

[00:54:02] interfacing with the

[00:54:03] customer the

[00:54:03] moment he sees

[00:54:04] I mean the vendor

[00:54:04] the moment he sees

[00:54:05] the email that comes

[00:54:06] from me he'll be

[00:54:07] like

[00:54:08] this is not

[00:54:08] he's asking so

[00:54:09] many things

[00:54:10] everything he's

[00:54:10] asking to

[00:54:11] leave alone after

[00:54:12] that you asking

[00:54:13] for customization

[00:54:15] right

[00:54:16] this is on one

[00:54:17] side

[00:54:17] next

[00:54:19] if you are

[00:54:20] bringing to the

[00:54:21] there is

[00:54:22] there is an angle

[00:54:23] of price point

[00:54:24] that you can

[00:54:24] achieve the cost

[00:54:25] of manufacturing

[00:54:25] which is the

[00:54:26] number of hours

[00:54:27] that will take

[00:54:27] for you to make

[00:54:28] that particular

[00:54:28] product

[00:54:28] it's not

[00:54:29] enough that you

[00:54:30] have made

[00:54:31] something that

[00:54:31] works

[00:54:32] right

[00:54:33] customer has

[00:54:34] set a benchmark

[00:54:35] of a price

[00:54:35] you have your

[00:54:36] BOM cost

[00:54:37] already

[00:54:37] there is a

[00:54:38] there is a

[00:54:39] leftover of

[00:54:40] margin within

[00:54:41] the money

[00:54:41] out of which

[00:54:42] you have to

[00:54:42] run your

[00:54:42] company

[00:54:43] which includes

[00:54:44] your R&D

[00:54:44] licenses

[00:54:45] your products

[00:54:46] capital

[00:54:46] sometimes capital

[00:54:47] investments

[00:54:48] predominantly

[00:54:48] operation

[00:54:49] cash cycle

[00:54:50] that you have

[00:54:51] to kind of

[00:54:51] make it

[00:54:52] work

[00:54:52] now imagine

[00:54:53] the

[00:54:53] imagine the

[00:54:55] layers of

[00:54:56] post sales

[00:54:57] pre sales

[00:54:58] interaction

[00:54:58] that you are

[00:54:58] doing too

[00:54:59] and plus

[00:55:00] how many people

[00:55:01] can produce

[00:55:02] are needed

[00:55:03] to produce

[00:55:04] one particular

[00:55:05] product

[00:55:05] how many

[00:55:05] hands is it

[00:55:06] going through

[00:55:06] how many

[00:55:06] manners

[00:55:07] does it

[00:55:08] require for

[00:55:08] you to

[00:55:08] produce this

[00:55:09] product

[00:55:09] if you don't

[00:55:10] include that

[00:55:10] into your

[00:55:12] into your

[00:55:12] this is

[00:55:13] applicable only

[00:55:14] for hardware

[00:55:14] design

[00:55:14] software is

[00:55:15] all this

[00:55:16] complexity of

[00:55:17] business is

[00:55:17] kind of

[00:55:18] gone in your

[00:55:19] software side

[00:55:20] of things

[00:55:20] cost of

[00:55:21] bits and bytes

[00:55:21] is zero

[00:55:22] zero

[00:55:22] exactly

[00:55:23] right

[00:55:23] and there is

[00:55:24] no effort

[00:55:25] to make the

[00:55:25] bits and

[00:55:25] bytes

[00:55:26] once you have

[00:55:26] made it

[00:55:27] to replicate

[00:55:27] the bits

[00:55:28] and bytes

[00:55:28] here to

[00:55:29] replicate the bits

[00:55:29] and bytes

[00:55:30] the nuts

[00:55:31] and bolts

[00:55:32] right

[00:55:32] is expensive

[00:55:33] you need to

[00:55:34] put a person

[00:55:35] full time

[00:55:35] right

[00:55:36] you need to

[00:55:36] make an

[00:55:37] infrastructure

[00:55:37] around it

[00:55:38] you need to

[00:55:38] shrink the

[00:55:39] time that

[00:55:39] it takes

[00:55:39] for him

[00:55:40] to produce

[00:55:40] that part

[00:55:42] right

[00:55:42] and he

[00:55:43] should be

[00:55:44] imagine

[00:55:45] if you're

[00:55:46] the car

[00:55:47] that you

[00:55:47] want to

[00:55:47] sit inside

[00:55:48] how safe

[00:55:49] do you

[00:55:49] want it

[00:55:49] to be

[00:55:51] right

[00:55:51] the weather

[00:55:52] conditions

[00:55:53] the road

[00:55:53] conditions

[00:55:53] a ton of

[00:55:54] material

[00:55:55] vibrating

[00:55:55] on top

[00:55:56] of that

[00:55:56] road

[00:55:57] that the

[00:55:57] wheel

[00:55:57] not running

[00:55:58] out

[00:55:58] on both

[00:55:58] sides

[00:55:59] right

[00:55:59] so

[00:55:59] it's not

[00:56:00] slipping on

[00:56:01] your rain

[00:56:01] and sun

[00:56:03] and so

[00:56:03] and so

[00:56:03] forth

[00:56:03] right

[00:56:04] so

[00:56:04] you want

[00:56:05] that

[00:56:05] and for

[00:56:06] next five

[00:56:06] years

[00:56:06] at least

[00:56:07] they don't

[00:56:08] have issues

[00:56:08] with this

[00:56:08] right

[00:56:09] that's the same

[00:56:10] expectation

[00:56:10] the customer

[00:56:11] also has

[00:56:11] there is

[00:56:11] wear and tear

[00:56:12] there is

[00:56:12] all of those

[00:56:13] things

[00:56:13] what if

[00:56:14] the guy

[00:56:15] has made a

[00:56:16] mistake

[00:56:16] in the way

[00:56:16] he connects

[00:56:17] the wire

[00:56:19] metal gets

[00:56:20] damaged

[00:56:21] you have that

[00:56:22] problem

[00:56:22] the kind of

[00:56:23] poke yoke

[00:56:23] that you have

[00:56:24] to think

[00:56:24] for each

[00:56:24] of those

[00:56:25] layers

[00:56:25] and structures

[00:56:26] where he

[00:56:26] will not

[00:56:26] make mistakes

[00:56:27] of what

[00:56:28] he's

[00:56:28] been

[00:56:28] you can't

[00:56:29] put an

[00:56:29] engineer to

[00:56:29] assemble these

[00:56:30] products

[00:56:31] you need

[00:56:32] anyone to

[00:56:32] be

[00:56:32] you need

[00:56:33] to think

[00:56:34] about automating

[00:56:35] through process

[00:56:35] first

[00:56:37] that's another

[00:56:38] another insight

[00:56:39] that I gained

[00:56:40] which we have

[00:56:40] to think

[00:56:41] about is

[00:56:41] automation

[00:56:42] doesn't happen

[00:56:43] with missions

[00:56:43] automation

[00:56:44] happens with

[00:56:44] people in

[00:56:45] process first

[00:56:46] where you

[00:56:47] could put

[00:56:47] anybody

[00:56:48] give an

[00:56:48] instruction

[00:56:49] set

[00:56:49] anyone

[00:56:50] can interpret

[00:56:50] the instruction

[00:56:51] set

[00:56:51] could actually

[00:56:52] repeat it

[00:56:53] that you're

[00:56:54] not dependent

[00:56:54] on one

[00:56:55] person

[00:56:56] if your

[00:56:56] technology

[00:56:57] demands you

[00:56:57] to be there

[00:56:58] to make

[00:56:58] it work

[00:56:58] they have

[00:56:59] not built

[00:56:59] a productization

[00:57:00] yet

[00:57:02] you have

[00:57:03] not built

[00:57:03] something usable

[00:57:03] in the first

[00:57:04] place

[00:57:06] so factoring

[00:57:07] in where

[00:57:08] you can

[00:57:09] have

[00:57:10] people

[00:57:11] for

[00:57:11] manufacturing

[00:57:13] where you

[00:57:14] don't need

[00:57:14] a highly

[00:57:14] skilled person

[00:57:15] to put

[00:57:15] your

[00:57:16] product

[00:57:16] together

[00:57:16] so have

[00:57:17] you thought

[00:57:17] that in

[00:57:18] your

[00:57:19] in your

[00:57:20] in the way

[00:57:21] you're putting

[00:57:21] designing your

[00:57:22] hardware

[00:57:22] how would

[00:57:23] you be able

[00:57:24] to make

[00:57:24] them

[00:57:24] how would

[00:57:24] you be able

[00:57:25] to mention

[00:57:25] them

[00:57:25] you know

[00:57:26] that's one

[00:57:26] of the first

[00:57:27] things that

[00:57:27] we even

[00:57:27] put as

[00:57:28] problem

[00:57:29] statements

[00:57:29] for

[00:57:29] people

[00:57:30] whom we

[00:57:30] recruit

[00:57:31] right

[00:57:33] we say

[00:57:34] is it

[00:57:34] manufacturable

[00:57:35] so your

[00:57:36] design

[00:57:36] functionally

[00:57:36] functional

[00:57:37] design

[00:57:37] is given

[00:57:37] for nobody

[00:57:38] is going

[00:57:38] to pay

[00:57:38] you

[00:57:38] if it's

[00:57:39] not even

[00:57:39] working

[00:57:40] first of

[00:57:40] all

[00:57:40] but getting

[00:57:41] it working

[00:57:41] just doing

[00:57:42] the

[00:57:42] particular

[00:57:42] task

[00:57:43] it's

[00:57:43] able to

[00:57:43] pick

[00:57:43] and then

[00:57:44] drop

[00:57:46] have

[00:57:47] you

[00:57:47] designed

[00:57:47] in such

[00:57:47] a way

[00:57:47] that it's

[00:57:48] manufacturable

[00:57:48] have

[00:57:49] you

[00:57:49] designed

[00:57:49] them

[00:57:49] in the

[00:57:50] way

[00:57:50] that it

[00:57:50] is

[00:57:50] serviceable

[00:57:52] have

[00:57:52] you

[00:57:52] designed

[00:57:53] it

[00:57:53] in

[00:57:53] the

[00:57:53] way

[00:57:53] that it

[00:57:53] is

[00:57:53] scalable

[00:57:54] and modular

[00:57:56] right

[00:57:57] so that

[00:57:57] something

[00:57:58] is

[00:57:58] primary

[00:57:59] and they

[00:57:59] keep

[00:58:00] missing

[00:58:00] this

[00:58:00] every

[00:58:01] time

[00:58:01] and this

[00:58:02] is

[00:58:02] not a

[00:58:02] thought

[00:58:02] process

[00:58:02] that

[00:58:03] comes

[00:58:04] right

[00:58:04] because

[00:58:04] it's

[00:58:04] not just

[00:58:04] about the

[00:58:05] manufacturability

[00:58:06] many times

[00:58:06] you get

[00:58:07] the piece

[00:58:08] machined

[00:58:08] in a certain

[00:58:08] way

[00:58:09] you say

[00:58:09] this

[00:58:09] shape

[00:58:10] is

[00:58:10] possible

[00:58:11] but

[00:58:11] actually

[00:58:12] he will

[00:58:17] my camera

[00:58:18] needs

[00:58:18] a 20

[00:58:19] micron

[00:58:19] of

[00:58:19] assembly

[00:58:20] tolerance

[00:58:21] that's

[00:58:22] like

[00:58:22] one-fifth

[00:58:23] of a

[00:58:23] hair's

[00:58:23] thickness

[00:58:23] is the

[00:58:24] deviation

[00:58:24] that it

[00:58:24] can have

[00:58:25] between

[00:58:25] two

[00:58:25] eyes

[00:58:26] the

[00:58:26] more

[00:58:26] I

[00:58:27] increase

[00:58:27] this

[00:58:27] the

[00:58:28] error

[00:58:29] that

[00:58:29] happens

[00:58:29] on this

[00:58:29] between

[00:58:30] the

[00:58:30] camera

[00:58:31] in terms

[00:58:31] of

[00:58:31] angle

[00:58:32] 81

[00:58:33] times

[00:58:33] of

[00:58:33] processing

[00:58:33] keeps

[00:58:34] increasing

[00:58:36] right

[00:58:36] now

[00:58:37] if I

[00:58:38] keep

[00:58:38] increasing

[00:58:39] this

[00:58:39] 81

[00:58:39] times

[00:58:39] of

[00:58:39] processing

[00:58:40] every

[00:58:40] time

[00:58:40] right

[00:58:41] let's

[00:58:46] need

[00:58:47] a

[00:58:47] processor

[00:58:47] I

[00:58:48] need

[00:58:48] to

[00:58:48] achieve

[00:58:48] the

[00:58:48] task

[00:58:49] within

[00:58:49] the

[00:58:49] same

[00:58:49] time

[00:58:49] within

[00:58:50] that

[00:58:50] one

[00:58:50] second

[00:58:50] whatever

[00:58:51] it's

[00:58:51] supposed

[00:58:51] to

[00:58:51] do

[00:58:52] now

[00:58:53] I

[00:58:53] need

[00:58:54] a

[00:58:54] processor

[00:58:55] that is

[00:58:55] 81

[00:58:55] times

[00:58:55] more

[00:58:56] powerful

[00:58:56] to

[00:59:16] greasy

[00:59:17] dusty

[00:59:18] oily

[00:59:18] and

[00:59:19] hot

[00:59:21] so

[00:59:21] your

[00:59:21] fan

[00:59:22] will

[00:59:22] start

[00:59:22] accumulating

[00:59:23] all

[00:59:23] those

[00:59:24] gunk

[00:59:24] on top

[00:59:25] of it

[00:59:26] it's

[00:59:26] going

[00:59:26] to

[00:59:27] slow

[00:59:27] down

[00:59:30] then

[00:59:32] the

[00:59:32] effective

[00:59:32] heat

[00:59:33] removal

[00:59:33] is

[00:59:33] reduced

[00:59:33] then

[00:59:34] the

[00:59:34] processor

[00:59:35] clock

[00:59:35] is

[00:59:36] downgraded

[00:59:37] then

[00:59:37] the

[00:59:38] timing

[00:59:38] it is

[00:59:38] supposed

[00:59:38] to

[00:59:39] achieve

[00:59:39] is

[00:59:46] this

[00:59:47] is

[00:59:47] one

[00:59:47] example

[00:59:47] it

[00:59:48] goes

[00:59:48] all

[00:59:48] the

[00:59:49] way

[00:59:49] down

[00:59:49] there

[00:59:49] second

[00:59:51] imagine

[00:59:51] the

[00:59:51] cost

[00:59:52] of

[00:59:52] never

[00:59:52] think

[00:59:53] that

[00:59:53] it's

[00:59:53] going

[00:59:53] to

[00:59:53] work

[00:59:53] absolutely

[00:59:54] fine

[00:59:54] your

[00:59:55] product

[00:59:55] is

[00:59:55] going

[00:59:55] to

[00:59:55] fail

[00:59:56] for

[00:59:57] sure

[00:59:57] in fact

[00:59:57] I

[00:59:58] say

[00:59:58] the

[00:59:58] statement

[00:59:58] saying

[00:59:59] you

[01:00:00] know

[01:00:00] when

[01:00:00] your

[01:00:01] product

[01:00:01] is

[01:00:01] going

[01:00:01] works

[01:00:02] only

[01:00:02] when

[01:00:03] your

[01:00:03] product

[01:00:03] fails

[01:00:04] right

[01:00:05] because

[01:00:05] you

[01:00:06] figure

[01:00:06] the

[01:00:06] boundary

[01:00:06] conditions

[01:00:07] of

[01:00:08] after

[01:00:08] the

[01:00:08] product

[01:00:09] fails

[01:00:09] is

[01:00:09] the

[01:00:09] customer

[01:00:10] going

[01:00:10] to

[01:00:10] come

[01:00:10] back

[01:00:10] to

[01:00:10] buy

[01:00:11] it

[01:00:11] from

[01:00:11] you

[01:00:11] or

[01:00:11] is

[01:00:12] going

[01:00:12] to

[01:00:12] move

[01:00:13] on

[01:00:13] to

[01:00:13] somebody

[01:00:14] else

[01:00:14] or

[01:00:14] not

[01:00:14] use

[01:00:14] the

[01:00:15] product

[01:00:15] is

[01:00:15] significant

[01:00:15] value

[01:00:16] that

[01:00:17] there

[01:00:17] is

[01:00:17] some

[01:00:17] tolerance

[01:00:18] to

[01:00:18] failure

[01:00:18] failure

[01:00:19] right

[01:00:20] like

[01:00:20] is it

[01:00:21] failing

[01:00:21] on

[01:00:21] the

[01:00:21] same

[01:00:21] day

[01:00:21] that

[01:00:22] he

[01:00:22] bought

[01:00:22] is

[01:00:22] it

[01:00:22] failing

[01:00:22] five

[01:00:22] years

[01:00:23] down

[01:00:23] the

[01:00:23] line

[01:00:23] and

[01:00:23] then

[01:00:23] after

[01:00:24] it

[01:00:24] fails

[01:00:24] is

[01:00:24] it

[01:00:24] coming

[01:00:25] only

[01:00:25] when

[01:00:25] it

[01:00:25] comes

[01:00:25] back

[01:00:26] you

[01:00:26] have

[01:00:27] your

[01:00:27] product

[01:00:27] market

[01:00:27] fit

[01:00:29] all

[01:00:29] this

[01:00:30] PMF

[01:00:31] statements

[01:00:31] that

[01:00:32] they

[01:00:32] talk

[01:00:32] about

[01:00:33] in

[01:00:34] the

[01:00:34] air

[01:00:34] that

[01:00:35] you

[01:00:35] are

[01:00:35] borrowing

[01:00:35] some

[01:00:36] when

[01:00:37] you

[01:00:37] translate

[01:00:38] that

[01:00:38] to

[01:00:38] implementability

[01:00:39] it

[01:00:39] is

[01:00:40] easy

[01:00:40] to

[01:00:40] throw

[01:00:40] those

[01:00:40] terms

[01:00:41] right

[01:00:41] price

[01:00:42] market

[01:00:42] fit

[01:00:43] is

[01:00:54] going

[01:00:54] to

[01:00:54] work

[01:00:54] there

[01:00:55] tech

[01:00:56] market

[01:00:57] fit

[01:00:57] is

[01:00:57] very

[01:00:57] different

[01:00:57] from

[01:00:58] so

[01:00:58] there

[01:00:59] are

[01:00:59] these

[01:00:59] layers

[01:00:59] that

[01:00:59] are

[01:01:00] missing

[01:01:00] yeah

[01:01:00] what

[01:01:01] I'm

[01:01:01] kind

[01:01:02] of

[01:01:02] imagining

[01:01:02] is

[01:01:02] everything

[01:01:03] is a

[01:01:03] component

[01:01:04] in

[01:01:04] this

[01:01:04] mega

[01:01:04] chain

[01:01:05] yes

[01:01:05] right

[01:01:06] and

[01:01:06] if

[01:01:07] you

[01:01:07] increase

[01:01:08] cost

[01:01:09] I mean

[01:01:10] it is

[01:01:11] not just

[01:01:12] impacting

[01:01:12] your

[01:01:12] customer

[01:01:12] but

[01:01:13] your

[01:01:13] customers

[01:01:13] is

[01:01:13] customers

[01:01:14] it's

[01:01:15] a

[01:01:15] chain

[01:01:15] it's

[01:01:15] a

[01:01:15] chain

[01:01:16] it's

[01:01:17] loaded

[01:01:17] yeah

[01:01:18] right

[01:01:18] and

[01:01:19] depends

[01:01:19] on

[01:01:19] which

[01:01:19] layer

[01:01:20] in

[01:01:20] the

[01:01:20] industry

[01:01:21] are

[01:01:21] you

[01:01:21] hitting

[01:01:21] are you

[01:01:21] hitting

[01:01:22] just a

[01:01:22] step

[01:01:22] before

[01:01:23] the

[01:01:23] customer

[01:01:23] or

[01:01:23] are

[01:01:24] you

[01:01:24] hitting

[01:01:25] at

[01:01:25] the

[01:01:25] bottom

[01:01:25] layer

[01:01:25] and

[01:01:26] for

[01:01:26] a

[01:01:27] problem

[01:01:27] like

[01:01:27] this

[01:01:27] you

[01:01:27] don't

[01:01:28] know

[01:01:28] where

[01:01:28] all

[01:01:28] you

[01:01:28] would

[01:01:28] be

[01:01:28] hitting

[01:01:29] the

[01:01:29] tech

[01:01:30] like

[01:01:30] this

[01:01:30] where

[01:01:30] human

[01:01:31] hand

[01:01:31] is

[01:01:31] involved

[01:01:31] in

[01:01:31] all

[01:01:31] these

[01:01:31] layers

[01:01:32] so

[01:01:33] everywhere

[01:01:33] there's

[01:01:33] a

[01:01:33] potential

[01:01:34] for

[01:01:34] your

[01:01:34] robot

[01:01:34] to

[01:01:34] come

[01:01:35] and

[01:01:35] play

[01:01:35] right

[01:01:36] so

[01:01:37] there

[01:01:37] comes

[01:01:37] that

[01:01:38] understanding

[01:01:38] again

[01:01:39] right

[01:01:39] so

[01:01:39] if

[01:01:40] it's

[01:01:40] going

[01:01:40] to

[01:01:40] fail

[01:01:41] then

[01:01:42] it's

[01:01:42] going

[01:01:42] to

[01:01:43] cost

[01:01:43] every

[01:01:43] time

[01:01:43] it

[01:02:01] without

[01:02:01] my

[01:02:02] production

[01:02:02] being

[01:02:02] interrupted

[01:02:03] can I

[01:02:03] continuously

[01:02:04] think

[01:02:05] on those

[01:02:05] lines

[01:02:05] there's

[01:02:06] a

[01:02:06] possibility

[01:02:07] that

[01:02:07] you

[01:02:07] could

[01:02:07] put

[01:02:07] a

[01:02:07] service

[01:02:07] engineer

[01:02:08] cheaper

[01:02:08] than

[01:02:08] having

[01:02:10] do it

[01:02:10] charge

[01:02:11] on

[01:02:11] that

[01:02:12] and

[01:02:12] if

[01:02:12] it's

[01:02:12] profitable

[01:02:13] for the

[01:02:13] customer

[01:02:13] on and

[01:02:13] above

[01:02:13] that

[01:02:14] great

[01:02:15] right

[01:02:16] there are

[01:02:16] people

[01:02:16] who

[01:02:16] actually

[01:02:17] supply

[01:02:17] machines

[01:02:17] and

[01:02:17] along

[01:02:18] with

[01:02:18] that

[01:02:18] machine

[01:02:19] operator

[01:02:19] that

[01:02:19] they

[01:02:19] charge

[01:02:19] per piece

[01:02:20] this much

[01:02:21] and so

[01:02:22] on and so

[01:02:22] forth

[01:02:23] even in

[01:02:23] India

[01:02:24] like

[01:02:24] MPI

[01:02:24] which is

[01:02:25] magnetic

[01:02:25] particle

[01:02:25] inspection

[01:02:26] that

[01:02:26] they do

[01:02:26] for your

[01:02:27] forged

[01:02:28] parts

[01:02:28] and

[01:02:29] casted

[01:02:29] parts

[01:02:29] often

[01:02:30] to operate

[01:02:31] the machine

[01:02:32] they also

[01:02:32] supply

[01:02:32] a trained

[01:02:34] person

[01:02:34] along with

[01:02:35] it

[01:02:35] and the

[01:02:36] customer

[01:02:36] pays

[01:02:37] per piece

[01:02:38] cost

[01:02:38] to them

[01:02:40] in their

[01:02:40] facility

[01:02:41] sometimes

[01:02:42] those are

[01:02:42] also

[01:02:43] possible

[01:02:43] those models

[01:02:43] are also

[01:02:44] possible

[01:02:45] that's

[01:02:45] left to

[01:02:46] your imagination

[01:02:46] but factor

[01:02:47] this

[01:02:48] so

[01:02:50] again

[01:02:51] summarizing

[01:02:52] what you

[01:02:52] mentioned

[01:02:53] is that

[01:02:54] you have to

[01:02:54] think

[01:02:55] not just

[01:02:55] of the

[01:02:55] thing

[01:02:56] itself

[01:02:56] but how

[01:02:57] it fits

[01:02:57] into

[01:02:58] this

[01:02:58] whole

[01:02:58] chain

[01:02:59] of

[01:03:00] components

[01:03:01] right

[01:03:02] and you

[01:03:02] have to

[01:03:03] think

[01:03:03] in terms

[01:03:03] of

[01:03:03] systems

[01:03:04] and not

[01:03:04] just

[01:03:04] problem

[01:03:05] solution

[01:03:05] right

[01:03:07] and in fact

[01:03:08] one more

[01:03:08] point

[01:03:08] I missed

[01:03:09] when I

[01:03:10] was

[01:03:10] elucidating

[01:03:11] that

[01:03:11] 20 microns

[01:03:12] is how

[01:03:12] much

[01:03:12] it is

[01:03:12] important

[01:03:14] getting

[01:03:15] the 20

[01:03:15] microns

[01:03:15] alignment

[01:03:16] there is

[01:03:17] certain

[01:03:17] ways of

[01:03:18] cutting

[01:03:18] the

[01:03:19] part

[01:03:19] the

[01:03:20] material

[01:03:20] there are

[01:03:21] so many

[01:03:21] ways

[01:03:22] to remove

[01:03:22] material

[01:03:23] which is

[01:03:24] the way

[01:03:24] you should

[01:03:24] adopt

[01:03:26] so that

[01:03:26] you can

[01:03:27] get the

[01:03:27] consistency

[01:03:28] of the

[01:03:28] tolerance

[01:03:28] on both

[01:03:29] sides

[01:03:29] if that

[01:03:30] guy has

[01:03:30] to remove

[01:03:30] it and

[01:03:30] flip it

[01:03:31] and put

[01:03:31] it

[01:03:31] 20

[01:03:31] microns

[01:03:32] how do

[01:03:32] you ensure

[01:03:32] that he

[01:03:33] is exactly

[01:03:33] in the

[01:03:33] same

[01:03:34] he can

[01:03:35] place

[01:03:35] it

[01:03:35] right

[01:03:36] it's

[01:03:36] impossible

[01:03:37] right

[01:03:38] so then

[01:03:39] it's

[01:03:39] gone

[01:03:39] so you

[01:03:39] need to

[01:03:40] have

[01:03:40] it

[01:03:40] as

[01:03:40] one

[01:03:40] single

[01:03:41] shot

[01:03:41] can

[01:03:41] you

[01:03:41] cut

[01:03:41] it

[01:03:42] with

[01:03:42] the

[01:03:42] tolerances

[01:03:43] not

[01:03:43] changing

[01:03:43] so that

[01:03:44] then

[01:03:44] it

[01:03:44] changes

[01:03:45] the

[01:03:45] whole

[01:03:45] design

[01:03:45] for

[01:03:46] you

[01:03:46] right

[01:03:46] and

[01:03:47] it

[01:04:07] let's

[01:04:08] say

[01:04:08] your

[01:04:09] probability

[01:04:09] of your

[01:04:10] failure

[01:04:10] is very

[01:04:10] high

[01:04:11] you will

[01:04:12] end up

[01:04:12] having

[01:04:12] a stock

[01:04:12] always

[01:04:13] for every

[01:04:14] system

[01:04:14] that

[01:04:14] is