Why He Walked Away From Microsoft To Build India's AI | Rishi Bal, BharatGen CEO

[00:00:00] Today I would say some of the biggest bottleneck is getting our talent to be AI ready. We simply don't have the number of people with advanced degrees in machine learning and AI. We have to find a way to keep upskilling our talent base. So what's your advice to a 21-22 year old kid who's going to graduate? I think the most important thing that you can learn right now

[00:00:26] India stands the most to lose, you know, with the potential disruptions that AI can cause. I think India has to find the India model. Meet Rishi Bal, the man who came back from Microsoft to build something much bigger. India's own sovereign AI. Because the AI race is no longer between companies. It's between countries.

[00:00:53] Hey Rishi, thank you so much for being on Bharatvaarta. Pleasure meeting you and we're going to have an important conversation today. Thanks so much for having me. Yeah. So Rishi, why do we need sovereign AI and what does it mean actually? So maybe let's start with the definition of sovereignty. Sovereignty is a guarantee. A guarantee that you will always have access to a critical piece of technology.

[00:01:18] So whether we're talking about energy sovereignty or AI sovereignty, it's that guarantee that you will always be able to access a critical resource, right? Now sovereignty is also a few other things. Sovereignty is the ability to always have access, but also know what has gone into the AI. To also be able to, you know, control and edit the AI to make it work the way you want.

[00:01:43] So that if it's not working and it's 2am, you can make a phone call and you can service that AI to work in the way that you want. Right. And you think we can't get this level of reliability or predictability with any of these, you know, globally relevant models or, you know, the open source ones and so on? I think the open source models and the frontier models from other countries will have an open will have an important role to play.

[00:02:10] But at the end of the day, you want to have a thriving ecosystem locally that can actually meet your own needs. Otherwise, you're sort of waiting for somebody else to come and solve a problem when it's commercially interesting for them to come and solve it.

[00:02:25] So if AI is going to be as important as we think it will be in the future, then we shouldn't be waiting for somebody else to solve our AI needs in agriculture, in Ayurved, in legal, until it's commercially interesting. It needs to be something that we can control the roadmap on and we can control the behavior on. Right. But we don't have our own Internet. We don't even have our own social media, let's say.

[00:02:52] Right. So is there a point of like Indianizing some of these globally relevant models to the point where, you know, it's hard to distinguish between whether they are Indian or global or do you think there's always going to be a distinction? I love that question. And you're right. We don't have I think there's there's pieces of shared infrastructure and the Internet is one of those. But we do have our own textbooks and we do have our own TV channels. Right.

[00:03:20] And we do have our own media that publishes things from an Indian perspective. So beyond sovereignty as as guarantee to access, there's an aspect of Indianness that we have to think about. Right. Which is, does it speak our language? Does it understand our culture? And does it speak from a perspective that we share? Right.

[00:03:46] Because there's there's excellent. And to be clear, I think global media has a role to play, but national media does as well. And I think we're going to find the same in AI where globally, I will have its own role and and and local Indian AI will have its own. Right. And when we say sovereignty in AI, right, I mean, what does it actually mean and practice? Because, you know, it could be very complex. Right. I mean, the ecosystem itself is fairly large. Right. From all the way from chips to applications. So what does sovereignty mean?

[00:04:16] And you think that we should focus on, you know, sovereignty at every layer or is there like a critical piece of the stack that we should focus on? It's a it's a great question. I think you have to start somewhere. And I think a great place to start is that the is at the intelligence layer itself. Right. To say, you know, is to talk about sovereignty when you talk about the guarantee of access, let's talk about how you get to guarantee of access.

[00:04:44] Some pieces within AI are things that we will need to build ourselves. Some pieces are places where we can contribute into the open source community so that we and the rest of the world can build it together. And then there's some places where we will take tactical dependencies on particular pieces that are proprietary. Right.

[00:05:07] And the degree to which we want to participate in one versus the other is part of what, you know, we will figure out as part of this sovereignty of AI stack. So I think it's very important to have an entirely Indian local alternative to AI so that regardless of what public policy takes shape in another country, that you will always have access to your to your own AI.

[00:05:31] Right. Right. It'll be important to have infrastructure for inference and training that's open so that that access remains and that there's an opportunity where India can lead the way to create a software collective that other nations can participate in as well. And so this is something that we can build together. Right. And then finally, it is important to be able to have some degree of sovereignty in silicon as well.

[00:05:58] Now, whether that comes from Indian companies building it or whether it comes from partnerships with foreign companies, it's important to maintain some level of control across that entire stack. Right. So, you know, we don't manufacture our own chips. We don't have companies that do the GPUs either. Right. So we're attempting to build the infralayer. Right.

[00:06:20] But, you know, people say that a lot of this is going to get commoditized and perhaps the greatest value lies in building applications. Right. Build superb applications that are globally relevant, that have a lot of value. And perhaps, I mean, start from there. Right. What's your opinion? Yeah. I think as a as a country, that's what number four GDP today, soon to be number three.

[00:06:46] We have to sort of, you know, put on our big boy pants and look at the world in that in that context. And what I mean by that is, you know, let me maybe map this to the aircraft industry. It's important to have your own airlines and to have your own pilots. And that's absolutely something that we should create. And that's the application layer. Yeah. But if you don't get in on the ground floor on creating the aircraft engine itself. Right.

[00:07:14] Right. It's the ecosystem around the aircraft engine is so complex in terms of the material science, the electronics, the mechanics of it. That if you don't get in on the ground floor, it's very hard to come in and build that ecosystem two decades too late.

[00:07:32] Yeah. Right. And so it's important for us to be here at the ground floor as this technology develops so that we can not just learn to fly the plane and fly people in it, but also build the plane and build all the components necessary for it. Right. Right. So it's like that saying, right? The best time to plant a tree was 10 years back. The second best time is now. Right. I think you nailed it. Yeah. Yeah.

[00:07:59] You know, the US and China have gone about building AI in fundamentally different ways. Right. In the US, I mean, we've seen crazy amounts of venture capital money. And then, of course, with China, I mean, there's state sponsored capital and talent and so on and so forth. Right. I mean, could you talk about these two different ecosystems and, you know, what should we learn from these ecosystems? What should we avoid and so on?

[00:08:22] Right. These two ecosystems are so different. Right. And they're well suited for what each country does really well. Right. The sort of the entirely capitalist approach driven by venture money, sovereign funding, you know, profits that are being turned back into R&D for AI is the American model.

[00:08:45] Right. And it's been remarkably successful for what America does well. The Chinese model with their slightly more closed ecosystem of players and the government funding is remarkably successful for what China is doing. Right. Now you're seeing two completely different models turning out, cutting edge innovation, some of the latest papers across across the top journals in the world coming out from these two countries.

[00:09:11] I think India has to find the India model, the model that works well for us. And we have to go back and ask, what does the India playbook look like? Right. And where does the digital commons or the public rails, where does that fit in as part of our AI ecosystem and AI strategy? Right.

[00:09:32] How do we pull together resources to build the kinds of things that we need to in India? How do we take a global partnership approach with other nations that aspire to build their own AI so that this is something that we can actually do together, maybe support the open community in a different way?

[00:09:51] So I think there is a third way in addition to these two remarkably successful ways that India has to help co-create for ourselves and others in the world as well. So we've been super successful building out this large digital public infrastructure, right? I mean, whether it is Aadhar or UPI and so on. So do you kind of envision AI being like another typical stack that we can build in that sense?

[00:10:19] I think we should have sort of an AI commons, right? And what I mean by that is it's important for us to have a robust open source or open weights AI ecosystem that any startup or company in India can go and tap when they're building applications. It's also important to have a robust hosted infrastructure so that any person building an app can just go call an API and at a very low cost of inference, right?

[00:10:49] You can be able to reach that level of scale, national scale, drive prices down and provide an extremely low cost of inference that can drive adoption of AI. So I think both of those are important aspects of what we should be doing. Right. What are some India specific nuances? You know, I can think of languages perhaps, I mean, most obvious ones, but are there other things that are particularly India specific for which, you know, we have to factor in when we're building models and such?

[00:11:18] So if you're using most models today that we use that are from the US sound like they're from the West Coast of America, right? And I always like to say that I think that's great if you're from the West Coast of America. But I think the rest of the world is going to want models that are like them, that speak like them, right?

[00:11:43] And it's so nuanced in terms of, you know, how culture gets embedded into intelligence, right? You know, when we speak to each other, we'll often do mixed code conversations without even realizing. We swap between English and Hindi and we'll mix it up and we'll make English our own. Or there's cultural nuances to what we do that are just simply different. You know, we greet people.

[00:12:10] We greet people when they come into our house and everybody gets offered a glass of water, right? And you can go across the span of the country and across the span of like, you know, economic status. And you'll still get a glass of water when you go to somebody's house. That's culture. Yeah. And that culture is embedded into the intelligence.

[00:12:30] So it's really important for us to have not just support for languages and dialects, but the importance of culture and perspective being embedded into the intelligence that serves us. Right. And how do you get all of this data, right? Because I'm sure it's a messy problem. A lot of it is unstructured. Yeah. Perhaps in, you know, floating in the ether, right? Not really digital and so on. So that seems to be a problem in itself.

[00:12:56] I have to say in recent years, we've not been great about writing, especially not writing online in our own languages as much as we need to. Right. And that's part of the homework that's needed when you're building AI in India is really digging deep to be able to get those nuances. Right.

[00:13:18] I mean, it's little things like if you enter the average Indian kitchen, the number of words we have for the bharatan that are the utensils or the vessels that we cook in is fairly nuanced. Right. And, you know, anyone that works in the kitchen will recognize that, you know, depending on the shape, the function of the utensil changes and we have different names for them. Even names change when the materials change.

[00:13:47] Where do we go get that? Right. Where is the sort of canonical book that allows me to go and get the names of every bharatan in every language across India? It often doesn't exist. Right. And so part of the work that we need to do and why I think we need sort of this AI commons is we need to go do the work and collect that data. Somebody has to do the hard work.

[00:14:12] And that's part of what part of what we're doing is to actually do the work to collect that data that that is more representative of us. Right. Yeah. I think that's useful context to understand. Finally, you know, what Bharat Jan does. Could you talk about Bharat Jan? You know, what is the vision and where you are right now? Yeah, for sure. So Bharat Jan is a rather unusual company because we are. Can you call it a company? I mean, it's it seems like an initiative more than a company. It is. But you have to give it some sort of like legal legal shape.

[00:14:42] And so, you know, we've created a Section 8 or a nonprofit company around it. And it's super unusual in that, you know, we're building foundational models, but we're nonprofit. And we're housed inside of an educational institute here at IIT Bombay. Yeah. And we're partnered across, you know, eight different other top institutes around the country. And oh, by the way, we're entirely government funded. Right.

[00:15:10] And so, you know, you visit our office and it looks like just like any other startup, but it's got a very different feel to it. Right. There's a certain mission that we have at Bharat Jan around, you know, giving India AI sovereignty. Right. But we started this journey about a year and a half ago, and I'm so pleased at the sheer velocity of progress that we have had here. Right.

[00:15:38] We went from a place where when we started building this AI, people thought it couldn't be done. Right. And in fact, you know, while we were building it, Sam Altman visited Delhi and, you know, proudly got on the stage and said, you know, effectively, you leave the AI building to us and you can build the application. Right. Right. His trip this time looked a little bit different. Right. And I think that's, you know, people people ask, what's the response to Sam? I'm like, the response is in what we're doing.

[00:16:06] The response is in the work and actually showing that this work can be done in India. And in just 18 months, we've built a whole succession of text foundational models across 22 languages with a large, large-ish 17 billion parameter model. And then we're building our own models for speech and we're building our own vision models that understand documents. It's all happening all at once at incredible speed. Right.

[00:16:37] But when you're structured as a not-for-profit, right? I mean, how does this all work? I mean, because, you know, as a for-profit company, I mean, you understand, right? You optimize for the top line and then, you know, everyone sort of rose in that direction. But as a not-for-profit, I mean, doesn't it get like super distributed and stuff? I mean, how do you prioritize? Well, that's part of, you know, the mission, right? Is the part of the mission is we're not just interested in building models. We're really interested in growing the AI ecosystem.

[00:17:06] So I often say that we also build models, right? Which is we, you know, we're growing the talent ecosystem. We have over 100 interns, BTECs, MTECs, PhDs across these institutes that are interning with us. So we're churning out the next generation of AI builders right here from these top institutes. We are publishing papers. We're publishing in the top journals.

[00:17:34] We're publishing recipes on how to create models, data sets, benchmarks, you know, that actually allows the creation of AI in a positive direction. Then we're also creating models and we're also working with partners to build applications on top of it. So the ecosystem really matters to me. And building models is one part of what we do. Right. And are there some use cases that you see like becoming popular right now?

[00:18:02] You know, what's been remarkable is the level of excitement in India around AI. Like every sector, every vertical, so many citizens, you know, tell me about what their AI journey is. I was on a flight back from Delhi and a gentleman sitting next to me pulls out his chat GPT. And he's on this, whatever their super high end subscription is.

[00:18:29] And he's telling me all about how AI has transformed his journey, his political journey and career. And so it's remarkable to see how everybody's interested and how everybody's experimenting with AI. There's a few sectors that I'm personally super excited about. I think some of the greatest impact in India can be in the educational sector. Right.

[00:18:58] The idea of putting a personal tutor in the pocket of every student, no matter where they are in India, is the kind of possibility that we need to chase down. Right. Because it really equalizes opportunity. And so then you could be sitting in a tier three, tier four town. And, you know, as long as you've got a cell phone tower near you, you have access to the world's best education. So I think that's going to be something that's transformative.

[00:19:26] It's going to be pretty remarkable if we get it right in the legal sector, too. We have a huge pendency of cases. Right. And, you know, depending on who you're talking to, just unlocking our backlog of court cases could be like a huge GDP boost, you know, 15, 20, 30 percent, you know, just by unlocking that.

[00:19:47] And again, you know, we talk a lot about leapfrogging, but here's an opportunity to leapfrog, which is, you know, can we can we be an AI first legal system? Can we be an AI first education system? Yeah. No, that's it's a fantastic opportunity. Right. Right. I mean, India often does these leaps forward. Right. Like how we jump directly from PC to mobile. I mean, yeah. So are we going to solve education for all by building, you know, enough schools for everyone? Maybe not.

[00:20:14] Right. But with AI, I mean, we're going to make sure that everyone has a tutor of their own. And similarly, you know, all of these backlogs and so on with the with the court cases and so and stuff. Right. And and about 40 percent of these are all land records and those kind of things which are bounded problems. Right. We can certainly solve those with code as well. Right. Right. You know, you mentioned that Sam Altman comment. And I remember, you know, the there it was kind of a shot in the arm for everyone. Right. It riled up a lot of people.

[00:20:41] But at the same time, people were also thinking that this is a very difficult problem. Right. I mean, we don't have the compute. We didn't think that we had the talent or resources and so on and so forth. Now, you know, sure, we may not require all of the, you know, tens of billions that are thereabouts that, you know, open AI has raised. And maybe it's going to cost us like a few hundreds of millions like, you know, deep seek and so on. But still, it's going to it's going to require a lot of cash.

[00:21:11] Right. And it's going to require this very finite skill set as well. All of this to come together and kind of solve this problem. So as of now, right, in 2026, what is the hardest problem to solve? So I would say that, look, I mean, going back to sort of that original premise, sort of every every long journey starts with the first step.

[00:21:34] Right. And I feel like, you know, we're in that early, early innings, you know, in terms of solving for the things that we need to solve. We've been very smart in India about riding the AI journey at the right level, which is that AI and the ability to make AI is depreciating quite quickly. Right. And so you could be building a frontier model.

[00:22:00] Minus N at a tiny fraction of the cost that it took to build frontier model minus N, you know, maybe a year ago or or two years ago. So I think we have to be very smart about, you know, picking our right operating points, you know, in terms of where we follow. I'd say today some of the biggest challenges we have. And by the way, I would have said GPU is the biggest challenge.

[00:22:25] But thanks to the AI mission and the work that the government has put in, you know, they're really helping us solve some of that GPU problem. Policy is a part of it. Financing is a part of it. But today, I would say some of the biggest bottleneck is getting our our talent to be ready. Right. And, you know, we simply don't have the number of, you know, people with advanced degrees in machine learning and AI that are needed to be able to innovate effectively.

[00:22:55] So it's a matter of time. I think we get there. But, you know, we're still in the early phases of our journey. Right. Yeah. We have something like, I don't know, 10 or 12 like engineering graduates every year. Right. I mean, and there are these surveys yearly which come out saying that, you know, hardly 20 percent are employable or 30 percent are employable and so on.

[00:23:16] You think there is a way for us to get these folks to be productive and like maybe a good percentage of them, you know, work at places like Bharat Jain and affiliated companies and so on and build some frontier stuff? Yeah, I, you know, I think we have to do that. Right. We have to find a way to sort of keep upskilling our talent base. India stands the most to lose, you know, with the potential disruptions that that AI can cause.

[00:23:40] We really need to take a hard look at how we make people career ready and what role companies can have in terms of internships and connection between academia and industry that will allow students to be work ready. The other thing, though, I'll say is we have to find a way to sort of keep upskilling our talent base. India stands the most to lose, you know, with the potential disruptions that that AI can cause.

[00:24:08] Right. And so how do you enable the next generation of students that are graduating to be entrepreneurs? You know, AI gives you that big opportunity where you can go solve a problem that used to take six people, six months can now be solved by an individual in in a few weeks.

[00:24:25] AI gives people that opportunity. Now, how do we how do we sort of unlock that innovation and and allow our young graduates to be creators of companies, products and be able to, you know, support them as they scale that journey? I have a question around adoption, right, which seems to be a non-trivial problem for sure. Right. I mean, because I think as Indians, you know, we consume the highest amount of Internet per capita, whatever in the world.

[00:24:52] Right. And they're kind of used to getting the best things for for cheap. At least I mean, that's the aspiration and so on. So if you look at social media, we we we have tried before, but we haven't had a social media network that is indigenous, that is succeeded. Right. So do you think that we will have an AI model that is indigenous that will also succeed? I mean, do you see some challenges there in terms of getting adoption? You're right. You're right about sort of the history of that.

[00:25:21] And I'll maybe provide some background there is. I think we've had different sort of experiments with capitalism. Right. When I often joke that in the 70s, you had, you know, lots of options. You had two options of cars you could buy and two options of scooters. And that was and that was it. They were made in India. But, you know, the consumers really didn't have any choice. And it turned out we didn't innovate a lot.

[00:25:47] And then it feels like in the Internet era, we sort of opened it all up. And, you know, consumers got access to the best products for, quote unquote, free. But I don't think we've really sort of, you know, factored in the price of free. Right. What do we really mean? Right. Because nothing really is free. And let me give you an example.

[00:26:08] If somebody took, let's say, a ton of steel that took $100 billion or tons of steel that took $100 billion to produce and then put it into our market for free, we would stop that immediately. Right. We'd be like, hold on, that's price dumping. We don't allow that. But when the same thing happens in technology, we somehow seem to turn sort of a blind eye towards it where we're like, oh, well, it's free. This one is OK.

[00:26:36] $100 billion of these AI services are free. The thing that we have to be careful about free is what it can do to local innovation. Right. Which is, if you've got some of the biggest tech giants in the world subsidizing the delivery of these services, it effectively will kill local producers. And that's something that from a policy perspective, we need to be very careful about. But how do we enforce that? Right.

[00:27:02] I mean, because people might say this is market productionism and, you know, that, you know, this is unfair on people, on consumers and so on. And why should we use something that's inferior while the world is like using frontier stuff and so on? Right. I mean, how do we. And also given that this is such a big unlock as well. Right. Just in terms of access, productivity and so on. Yeah, you're right. It is it is a hard trade off to me. Right.

[00:27:28] Which is, you know, access to some of the best AI tools are going to be a big driver for GDP growth as well. So I don't think it's a it's an either or type of type of conversation. But we do need to find something that balances access with enablement and local ecosystem and local local production. I think it's a matter of time before, you know, in India, we produce things that match up to sort of frontier levels.

[00:27:54] But even today for specific functions, I think we can have very competitive AI models in in the market today. It's just the ability to subsidize delivery to a billion people that requires an enormous amount of money. Right. And that's a place where I think Indian AI companies will struggle. But if you think about UPI, right, for example, obviously, I mean, high class technology and so on. Right.

[00:28:23] But really, where it succeeded was getting that ecosystem together. Right. I mean, you had all the banks, then the tech and the government. Right. Or policymakers coming together to, you know, craft such a valuable sort of offering that people don't consider alternates anymore. Right. I mean, I think, you know, every month the number of transactions kind of grow. So is there an equivalent that we can do for AI?

[00:28:49] You know, if we can layer it with whatever digital public infrastructure we have right now and make it so compelling that, sure, the models by themselves may not be like frontier frontier. Right. But the end value that we derive from it is so strong that, you know, it's obvious to use it. I think this is exactly the direction that we need to explore in India, which is the digital rails or the AI rails for the country.

[00:29:16] And part of it is consumer, but part of it is business or government facing as well. If you just take an example of creating digital rails for, let's say, government call centers. Right. If you're able to create a powerful ecosystem there where now you've got that data cycle going, there's data coming in that's making the model better and the model gets better. The service gets gets better. That could be the place you start in terms of digital rails.

[00:29:46] And then that that can grow into a commercially competitive offering. Right. And so you could see how digital rails can be a powerful starting point. Another example could be things like banking, where perhaps you're evaluating transactions, but you want to be able to do do this across across banks.

[00:30:08] Right. So if you're able to create sort of a collective where there's a there's a neutral third party that can actually use AI to analyze bank transactions and be able to help prevent fraud, you may be able to do this better through this through the data actually coming together and create a better AI model than is possible by any private bank. Right. So those are all examples of how we can use the digital sort of AI infrastructure or rails to actually innovate.

[00:30:38] Right. And at that point, the solutions are world class. Right. That actually the people aren't using people aren't using UPI over over, say, Visa or Master because simply because it's cheaper. Right. I mean, UPI is a remarkable product. Right. And that's that's what we should aspire to. Right. Every time you're in the US and the, you know, the Starbucks barista gives you like all the chump change. Right. And, you know, you're instantly transported 10 years back of 10 years back.

[00:31:07] That's exactly right. I was, you know, a few years ago, I was at the I was at the 9-11 memorial, the museum, and there was a long line outside. I said, surely there's to be a better way to buy a ticket. And and so, you know, I was looking ahead and it was like a website. So I was like, great, you know, just take a photograph, QR code. And I mean, and then I go to the payments page and there's like fill in your credit card number and name. Yeah, sure. Random kiosk. Yeah, why not?

[00:31:38] Who does that? Where's the QR code? Is the question that you're always asking as an Indian. Yeah. Yeah. What are some of these ecosystem partnerships that you're evaluating? And could you give like some examples of how things have worked out for you? So, I mean, we talked a little bit about the education ecosystem. Right. As that's one way by which we're doing open research. So in order to do effective, effective research in AI, you have to be able to operate at scale.

[00:32:06] Right. And a lot of research universities just aren't going to have access at that scale. So the consortium that we've put together allows that intersection of engineering scale with some of the smartest, brightest, you know, faculty around the country to come together and do experiments at a level of scale and AI that aren't possible, you know, by either group, by them, by themselves. And that's an example of what, you know, ecosystem sort of play looks like.

[00:32:35] I'll give you another one. We do something in data where we discovered that there's a number of these amazing organizations in India that are digitizing our manuscripts, old books, documents, manuscripts, all of it. Now, a lot of that is just getting into image form, but we're not unlocking the true digital value. And so we noticed this and we said, hey, y'all are doing something amazing here. And many of them are nonprofits.

[00:33:02] And we said, we're doing this thing here where we've built some of the most amazing OCR pipelines that are composites of different models. Let's come together. Can I offer you a service to to convert these images into digital text? And if you're OK with it, allow us to keep a copy to create create better models. We're not going to sell the data. We're not we're not doing anything else. We just want to teach our models to speak Bengali better.

[00:33:30] Right. And a lot of people said, yeah, sure. Right. And so that's an example, again, of an ecosystem where people can come together and sort of like create a greater good through through that partnership. Right. You know, you interact with a lot of students and professors and so on. Right. I mean, do you think like the best and brightest here are sort of interested in building for Bharat per se? I think some of them are right.

[00:33:57] I often joke that of the hundred people that are interning with us every year, maybe 30 of them are going to go abroad, but 70 of them will be here and they'll be building building in India. So I think there is a degree of enthusiasm that I see. Right. Look, our talent, our young talent is second to no one. Right. You look at the graduates coming out of these IITs. They're they're world class, truly world class. They have the brights.

[00:34:26] They have the willingness to work hard and they're innovative, creative. What they sometimes lack is sort of a playground in which they can do cutting edge work. We're giving them that. Right. So for a lot of people, it's not just about the package. It's about that ability to do amazing cutting edge work and grow their skills and ability. And I think that's part of what happens when I talk about the ecosystem. Right. It's part of what happens is now you've got an amazing group of people you can work with.

[00:34:54] You can work on cutting edge stuff and you don't have to go to Europe or America anymore to do that work. Right. Yeah. I think that's the catch basically. Right. You know, we mentioned the government a few times. Right. I mean, what do you think the government is doing? Right. What do you think we should do better? I, you know, I talked about I talked a little bit about the AI mission. Right. And that ability to put together the, you know, funding, you know, for GPUs.

[00:35:24] They've also put together something called the AI coach, which is trying to sort of create a repository of of models and data into one place. So that makes it easier for the next generation of builders to be able to build as well. I think those are, you know, I think it's remarkable to see sort of the level of engagement that we have seen from there. Another thing I see happening right is there's a lot of interest in the state governments.

[00:35:49] Right. Everyone's asking the question, how are we using AI to sort of drive better citizen access, better governance? And so that's been really exciting in terms of things that the government could do better or, you know, we'd love to see. We'd love to continue to see policy adjustments that, you know, allow for a level playing field for AI innovators in India.

[00:36:17] That opportunity to, you know, create a level playing platform where, you know, we can compete equally as well. So when you talk to folks in the government, right, I mean, do you think they get it? Do you think that, you know, they get the seriousness of what AI can do for humanity and India in particular? Oh, yes. I have to say I've been very pleasantly surprised by the quality of conversations that we have had.

[00:36:47] Especially with, you know, in the central government, with the Department of Science and Technology, with Mighty that runs AI mission. There's a lot of people that are thinking deeply about what the opportunity and risk around AI is and thinking about how to support the AI ecosystem. And it's fabulous to see. Right. You know, I mean, it's a bit of a unique journey that you've had, right? I mean, you've been in the private sector for so long, right?

[00:37:16] You come back to India from the U.S. And you dive headlong into a government initiative, right? Now, these are words that are poles apart, right? Fast-moving, innovative versus, you know, slow bureaucratic processes and sort of stifled with a lot of paperwork per se, right? I mean, what was this transition like? It's been a learning journey. I'll say that.

[00:37:45] With probably at least a few awkward moments along the way. You know, it really starts with, you know, it really starts with the small things. Like, you know, when you arrive for a meeting in the government, there's a very strict seating protocol.

[00:38:07] And as somebody who's a sort of a West Coast startup technology person, there was definitely a learning curve to, you know, learn where I needed to sit or, you know, what the appropriate sort of behavior in these places was. But I mean, it's, you know, look, every culture and subculture has its own unique protocols, right? And you just have to sort of like sort of decode what the protocols are and what the incentives of all the individuals are, right?

[00:38:35] And then just sort of like work in this, you know, new model of things. What I will say is that the, you know, that said, the scale and impact, the scale of impact that you can get working with the government, right? With 1 billion people is just tremendous, right?

[00:38:57] And so sort of that extra bit of paperwork that you have to do and the protocols sort of unlocks access to say, how do I, can we have impact on the billion people? That's kind of interesting. Right. You know, we were talking before we started recording that there's a whole bunch of folks in the US right now or elsewhere in the West, and they're considering returning to India, right?

[00:39:21] Some of it forced by perhaps, I mean, what has happened over the last year or so with, you know, the political situation, the social situation and whatnot, right? What would you say to kind of encourage them to expedite the journey back home? Man, I would encourage them to come visit. Come visit some of these companies in Mumbai, Hyderabad, Bangalore, Delhi.

[00:39:46] There's a state of optimism amongst sort of young Indian builders that is remarkable. And there are a few places in the world where you can see the level of optimism and energy as you're seeing in many parts of India. And so I would say this is the time.

[00:40:06] This is the time to come back and be part of something big, innovative and growing at an incredibly fast pace. How does the startup ecosystem compare to, you know, maybe the ecosystems in the US, right? For example, one of the critique that we often hear is that, you know, people are building quote unquote frivolous things, right? That we don't invest in deep tech. We don't have true innovation here and so on. I don't quite agree with that.

[00:40:36] But, you know, what is your take? So I'll talk about it. I'll talk about two sides of what I see, right? I talked about sort of the sort of that amazing mindset that I'm seeing amongst, you know, college graduates and people who are starting to build here.

[00:40:50] I talked about the level of government support that we're seeing, including more recently, if you're following along with the work coming out of ANRF and RDF, where we're creating massive funds to invest in deep tech alongside the venture community. Right. This is not traditional government money. Right. This is actually more like VC money, where the government is investing, right? Investing alongside VC.

[00:41:18] There's a certain degree of patience for deep tech. So I think the opportunity and the money is really coming along. What I'll say we are missing is maybe two things. One, we're still growing the mindset, right? In terms of being innovators, change makers. I think there's still a degree of wanting to build the next sort of tactical business solution, right?

[00:41:45] It needs a different mindset to go after something with a longer horizon. So I think we're still building out that mindset and risk taking ability. The other thing I think we're missing is we're missing the fantastically big exits, right? Because, look, money and mindset will follow the big exits, right?

[00:42:06] And the Indian acquisition ecosystem, just in terms of the larger Indian companies acquiring the smaller ones, isn't quite at the same pace that we see in the US. So I think the big acquisitions, the big exits will draw in the right mindset and money as well. Right. So it's a matter of like a maturing ecosystem. So over a period of time, I think this could get sorted is what you're saying. I think so. I think so.

[00:42:36] But I mean, we still we clearly have our work to do, especially with some of the larger Indian companies being important draws in this ecosystem. So when we talk about AI, right, I mean, we talk about like the massive impact it can have on society and especially the young folks. Right. I mean, because even as I was, you know, coming here and I mentioned to some people that, you know, I'm going to be talking to you. The number one question they had was, you know, how is it going to impact jobs?

[00:43:05] Right. Or, you know, my kid is going to graduate in a year. You know, what skills are super important and so on. Right. Yeah, of course we do. Because because this is what happens at every family reunion and friends get together. So what's your advice to like a 21, 22 year old kid who's going to graduate? I think the most important thing that you can learn right now is the ability to learn.

[00:43:28] Right. I think in times of change, like we're about to enter, having a growth mindset and sort of that ability to continuously learn and change and adapt is probably the number one attribute that that we need. Right. The other thing is I would say is don't is approach this from not from a fear mindset, but from an opportunity mindset. Right. Because there's two ways to look at it.

[00:43:55] One is to look at it and say, look, I graduated with a degree in accounting and it looks like, you know, AI is now going to take away some fraction of accounting jobs. Another way to ask that to look approach the same situation is to say, how can I be part of the change that actually drives the adoption of AI in accounting? Here's the other reality. The the need for people to do amazing high quality work has never left, has never gone away in the history of mankind.

[00:44:24] Right. We seem to evolve to other things. Right. And we seem to evolve to jobs that didn't exist. The thing that the activity we're doing right now did not exist a few years ago. Right. There were no podcasters. There were no podcasts. And yet here we are, you know, performing this activity that is that is mainstream today. So I think being keen about what these new opportunities are and jumping into those is is a big part of what the youth need to do.

[00:44:54] I think the baseline always shifts forward. Right. I mean, we don't we don't we don't settle on a baseline that was relevant maybe 10 years back and say that, OK, these are all solved problems. We're only going to have these fraction of problems to solve. I mean, we keep shifting the baseline forward. Right. So people are going to get that much more productive. Right. Yeah. And and of course, I mean, there's going to be a transition period as with any technological leap. Right. But directionally, I mean, that's that's how it's going to be.

[00:45:23] That's how it's going to evolve. Right. Right. You know, to end things, I want to play out the bull case and the bear case. Right. In terms of like India and AI. So if you fast forward, let's say five years or 10 years, what is the worst thing that could happen for us? Like if we let's say, I mean, if we get everything wrong and what is the best thing that could happen for us? If, you know, let's say all the ducks are in a row and like, you know, we're we're able to play it out. Right.

[00:45:47] I mean, go with the bull case first, because that sort of comes naturally to me is I think we if we if we if we get this right, you know, India is a global leader with a global leader in AI and we've leapfrogged some of the most challenging aspects of, you know, society today. Right. And so we're able to make these leapfrogs in education, in the legal system, in governance, in in in citizen access where, you know,

[00:46:15] we're we're able to bring remarkable amounts of productivity, GDP growth in into the country. Right. And I think AI is a tool that can allow a lot of these a lot of these things to happen. I think a bad case actually is one where we don't where we don't embrace AI, both as producers and as consumers. And we actually get left behind. Right. And that's the situation where we rely on somebody else to be AI creators.

[00:46:44] And for us to be sort of fearful of AI and and let others sort of pass us by. Right. So I think we're very clearly going with that bull case. Right. I think I think that's the mindset I see in India. We're embracing this and I think we're headed the right way. No, it it's it's really amazing to note that the folks in the government are taking this stuff really seriously as well. Right. So to end things, you know, how should people get involved with Bharat Jan?

[00:47:12] How can they contribute or support the initiative? Thanks for asking. Right. We're always looking for people, people to pitch in as as a nonprofit. I think there's a lot of work to be done with AI in India. If you work in software today, the opportunity to create benchmarks that are India specific has never been greater. We're looking for partners to help us create better, richer, India centric data sets. I think there's an opportunity to do that. We're releasing models into the wild.

[00:47:42] We'd love for people to to use those models, give us feedback, critical, good, positive feedback. It's all very useful for us to create, create this and, you know, keep keep the ideas flowing back. We're here to build out the ecosystem and we're looking forward to partner with as many people as possible. Awesome. Thank you so much, Rishi. This was a really inspiring conversation, I would say. And thank you for coming back and thank you for helping us level up on AI.

[00:48:11] So fantastic. Thank you so much. For sure. Thank you.