He dropped out of PhD to build Google, Facebook, Square, DoorDash - Gokul Rajaram

The one and only Gokul Raja Ram.

The only work experience I've had in India is actually working at IISc for a summer and so dropped out after my master's joined Google which was at that point the only company that was hiring. In the aftermath of the.com bust there was almost no one at Google who had worked in advertising before. I was privileged to see that journey from zero to multiple billions of dollars. Went through some ups and downs but ended up at the end was acquired by Facebook. Help Facebook go public Coinbase Management Team Square Help the company go public there join DoorDash organization of DoorDash help DoorDash go public or help Coinbase Republic as a board member. Stock markets are weighing machines in the long term and voting machines in the short term. GPT5 is probably going to cost north of a billion dollars to train a Google, a Facebook, an Amazon, etc. You are not going to want to lose this race. I think most of Sam's time is almost sudd spent in raising capital at this point. There's $2.3 trillion that are being spent on various IT and business services. Hard for startups to break out in functional AI companies.

Hi everybody, Sanjay Swamy here again from primev Partners and for this next episode of our entrepreneur oriented podcast we have someone that I have been wanting to have on the show for for the better part of several years at least, certainly two years. But I'm thrilled to welcome the one and only Gokul Raja Ram. Gokul has been both a friend and an advisor, a mentor, someone we've all looked up to and also a partner. Amit has worked with Gokul in the past at at Google and someone we've all tracked over the years and got to know him reasonably well over the past few years. Done some co investments together and Gokul of course has been very well known both and is very well known in Silicon Valley notably for his roles at Google in the AdSense product at Facebook which is now Meta and the Ad Pro as Chief Product Officer at Square which is now Block I guess as well as several other things I could go on and on and also has been, you know, board member at Coinbase, prolific angel investor not just in the US and but also around the world and notably in India. So would love to first of all welcome you Gokul to our show and look forward to exchanging views on various topics that are dear to our hearts and to our listeners. Thank you.

Thank you Sanjay. It's great to be here and I've been a huge fan of prime and you and excited to finally get a chance to do this. Thank you for having me.

That's kind of you. So, Gokul, maybe we'll dive in to a little bit about your journey. You know, right from your days growing up here and you know, your journey in the U.S. maybe you could give us a little bit of a snapshot.

Yeah. I grew up near Delhi and I went to undergrad at IIT Kanpur. The only work experience I have had in India is actually working at IASC for a summer in Bangalore. I was doing some research for Professor Raja Raman who was the chair of the CS department at IITK many years ago and was a great experience and that wetted my appetite to do more research. So I went to the US for grad school, got joined the PhD program in computer science at UT Austin, but then after a couple of years realized I wanted to write production code and so dropped out after my masters to go become a software engineer at a startup company called Juno, which was very early ISP in the US in the mid-90s. It was one of the few companies in the New York ecosystem which was building ad software alongside DoubleClick back in the day. So I worked in order to monetize the free email and Internet service they provided. We need to build an ad system. So that was the monetization engine and I got exposed to building ads products fairly early and so after a few years of doing that, got my mba, but I landed up in Silicon Valley and joined Google, which was at that point the only company that I was hiring in the aftermath of the dot com bust in the early 2000s. And I was lucky enough because I had initially thought of joining telcos because I was really interested in networking from an infrastructure point of view, but they all rescinded my offers because they were all going through hiring fees and layoffs. So I landed at Google, which is my third choice, strangely enough, and just destiny, and was put on advertising because I had worked on ad products earlier. There was almost no one at Google who had worked in advertising before and ended up working on a product which later became called Google adsense and accounted for over half of Google's gross revenue at the time of going public, was privileged to see that journey from zero to a fairly large scale. I think it was multiple billions of dollars and after that started a company that basically after two and a half years again, interesting time. It was during the global financial crisis, went through some ups and downs but ended up at the end Was acquired by Facebook. Again, timing was right because Facebook was just starting its journey just in 2010. About two years of Facebook in public, but got to experience the growth of Facebook Ads. Led the Facebook ads product team and helped Facebook go public. A lot of good learnings there. And then worked at Square again. Really good experience was on the management team of Square. Helped the company go public there and then joined DoorDash as a managing of DoorDash helped DoorDash go public. Really interesting IPO. I call it the Zoom IPO because it was during COVID We went public in December of 2020 and so instead of going to the to New York like we did with Square with DoorDash the whole management team was on a zoom screen with the, with the Nasdaq people in the middle. And that's how we saw ourselves go go public. And that was a screenshot that even appeared in the, in the news channel. So it was a really interesting experience sitting in the living room right over there with a good background behind me. And then I also participated or helped Coinbase go public as a board member. And that was an interesting one is Coinbase was what is called a direct listing where it was a different kind of IPO than normal. Google was a Dutch auction back in the day. So IPO had some unique characteristics. DoorDash is a Zoom listing. Facebook is the first time that NASDAQ ever came to campus. They actually came down to Facebook campus and we rang the bell from the Facebook campus. Mark rang the bell. So each of them was unique experience and unique. Just learning of what the different types of ways. But all of them ultimately they say stock markets are weighing machines in the long term and voting machine in the short term, which means ultimately companies have to perform well over the long term. And so thankfully all these companies have performed well and have done well and stock market has responded appropriately. Now I am a full time investor hoping to follow in your footsteps and Prime Venture's footsteps and invest like we have done alongside a few companies. It's been a great experience working with you and hopefully get to do more of those things and support great entrepreneurs building legendary companies.

Wow, that's quite the background, Gokul. I mean any one of those topic or those experiences is probably a podcast in itself. But what are maybe you know, for entrepreneurs here sort of getting into perhaps growth stage where you know, several of the companies you work with, they're probably already at that stage. What are some common things across, you know, this journey that you know, have served you very well and the Companies very well and what are perhaps some unique things about each of the companies. Maybe if you a couple of anecdotes would be great, you could share.

Yeah, I think the growth stage is a really interesting point. I always think the early growth versus scale up. So there are three stages to any company. There's a startup stage which is the only goal is to find product market fit at that point and you want to keep burn as low as possible and be scrappy. The second stage is, I call it the early growth stage where you just, you have, you're trying to build repeatability and you're trying to figure out okay, was this product market fit an illusion or was it, is it a real thing? So you're trying to figure out the channel, you're trying to figure out a little bit of things on repeatability. And then finally there's scale up where you figure out repeatability. You're just adding capital and just trying to scale as fast as possible. That early growth stage is the most important thing because that's when you start adding some process. So I think the biggest thing where people get wrong is they either add too much process or too little process. I think that that secret sauce, the right thing to get right. What is the right amount of process you add so that you can run the machine but not devolve into chaos or not become too bureaucratic. So you've got to draw the fine line between over bureaucratic, over bureaucrac and chaos. And right there is that is that sweet spot, the Goldilocks zone. And I think process is in a few places. One, around people. I think hiring is probably the most important thing. It was super important. But the first 10 hires you have mostly from your network for the most part. Now it's the first time you're going outside your network to hire in this phase. So how do you put a process into place where you're just like with customers, you're top of the funnel, you're getting enough people top the funnel and you're getting people through the funnel to get to an offer stage in a systematic way. So you've got to put some process into place. Second, you've got to put process around goal setting. So I think now there's enough people that most of you can't fit into one room. People fit into multiple rooms or even multiple locations. And so how do you set goals? How do you align people around goals so each person knows what their goal is and how it fits into the broader whole of goal of the company? So okrs Start coming into play.

So things like okrs.

Yeah, but yeah, I think that's the second piece. The third piece is around communication. I think it's very important on an ongoing basis to figure out as a leader of a company and as leaders of different teams, how you're communicating. Things like all hands, things like potentially weekly emails, things like even staff meetings, those things didn't happen. What is this? Product reviews, weekly reviews. What is the set of communication processes you need in the company so that everyone is executing and orchestrating things together? And finally, people process beyond hiring. So I think you have a group of people. This is when people start thinking, okay, how's my career going to evolve in this company? And so you start having to put in some lightweight performance plans, performance reviews, figuring out what ladders look like, compensation structure. So people have been again, during the PMF phase, no one's focused on, okay, what's my cash count going to be? It's like die or get to pmf. Right? That's the only thing. But once you get to pmf, people are okay. Now you raised maybe another round of funding, okay, you know, do I stay here for two more years? Am I going to become a senior software engineer from software engineer, what's my path look like, etc. Etc. And I think again, it's possible to overdo it, it's possible to underdo it for each of these things. The trick is figuring out how to not have too much bureaucracy. So you create 20 layers of a career ladder or you don't get any layers. The answer probably is for engineering you get three layers, you have a junior engineer layer step in the ladder, you have a mid level engineer and you have a senior engineer based on experience, based on contributions, based on scope, impact, etc, and that's it. And then maybe the next stage you have a little bit more ladder. So again, the goal is to try to do the minimum possible thing so that you can still. The goal is to find repeatability so everyone is on the same page, executing, marching to the same tune versus doing their own thing. And the company just dies.

Yeah. So, you know, I think one of the challenges I've noticed at least is companies where product market fit itself is not well defined. Right. And what is product market fit? Is it people using the product? Is it people paying for the product? Is it people coming back and repeating, you know, their subscriptions? You know, we've had interesting debates with SAS companies where they say, well, the customer is prepaid for a year. It's not always sure that that's a good thing because it may mean that at the end of the year they may churn out and they've stopped using it. And, and obviously you need to see how they're engaging versus perhaps even giving them the option to just go and leave monthly in the beginning so that they're only staying in the system if they're getting value. And what is, what is in the many of the companies that you work with also. Right. I mean product market fit was probably established in a few dimensions, but perhaps not quite in the monetization side of things. Right. And of course, as the world has evolved and in the post 2021 era, it's all about monetization these days and everybody's looking for almost profitability but certainly monetization and repeatable capability there. So for early stage founders, you know, again in the, in the PMF zone, we'll start and then come to the other two. What are your views on the importance of monetization and establishing unit economics and things like that?

I think that's a great question, Sanjay. Product market fit has been defined in numerous ways. Like you said, there's no, I would say there's no one clear definition. The way I think of it is product market fit has two parts. One I call problem solution fit, which is basically, basically can you on one side consistently when you get a new customer, can you consistently get them to see value in what you're offering? And second, for the first time, so can you activate them on a consistent basis? Second, can you retain them once they, once they see value, do they stick around? So those two together, activation and retention together, I call problem solution fit. So are you solving a clear problem and are you solving a clear problem on a consistent basis? So there the metric is around retention. So you want to see, you know, activation, retention together. In some ways there's also, you start looking at can you acquire customers efficiently? But efficiency is not as important as, because that goes into the cost piece of things. Can you acquire a customer in the first place, show them value and get them to retain. Let's just focus on that. So problem solution, are you solving a problem that they have and can you get them to understand this? So activate and then retain. The second piece is I think can you, can you be this, this I call go to market fit. And I think this is important because I think if we just stop there, it basically doesn't, doesn't really show whether or not you can get enough of these customers do the efficiently, etc. So first part of go to market fit is can you actually consistently get in front of your ideal customers? So can you reach them in an efficient way? And then can you consistently close deals with your ideal customer profile? So can you then convert them? So can you basically reach them and then convert them? And I think that together starts to think about, starts to get the channel into place. I do. So again, problem solution fit, which I think is the most important one you've got to find out. And the next part is you're trying to build repeatability on the go to market front. Now where does monetization fit in? I think it's a very good question. The challenge is, I think for companies we know for certain kinds of things, monetization can be deferred. For example, if you have a media company or a company that relies on consumer engagement, we know time and again this has been proven that if you can get enough engagement, enough engaged users, you can use advertising to monetize. And every time a new new engaging media came over, people said that can't monetize. Turns out it did monetize, whether it was Snapchat, Pinterest, Twitter, TikTok, all of them have figured out how to monetize using ads. So I think for consumer media properties, but you need massive scale. You need massive scale. You need at least 100 million MAUs, probably more, hundreds of million maus to build a reasonable business. For almost everything else where the value exchange is not in media time or attention, but it is in paying for a service, I think we've got to figure out our customers willing to pay. So willingness to pay is I think a very important part of the problem solution fit, which I think it has to be included because activation can't just be free activation. I mean it could be, but retention probably does involve because the freemium product maybe so you could activate with a free product, but retention, you got to test whether they're willing to pay or not. Why? Because if they are free, you don't really get a signal of what the real customer behavior is going to look like. And so I don't trust. So that's why you've got to. If you have a business model which relies on exchange of value through customers paying you for value, you've got to test it out because otherwise I don't believe the customer behavior will hold true when you introduce payments. So I do think it's important to fully simulate or fully test PMF with the right business model that you're going to have. Because if you say, oh, I have free Users. Now I'm going to introduce payments. I have a few companies in my portfolio that try to basically keep services that you otherwise would think of as paying services free for a very long period of time thinking oh will monetize later and you could with things like take rates etc but it becomes very, very, very challenging especially over the last few years. I think investors have refused to believe one's hypothesis around this and they, they, they basically are much more conservative than in entrepreneurs are in figuring out what the attach rate is of payments. So I think it's very important for entrepreneurs to take it this otherwise it'll be to their detriment if they don't.

Got it. Now it's also interesting that one of the things we experienced trying to not name the company but I'll have to give the example. Yeah where you know, they wanted to offer a service for 50 rupees per user per month and you know, typically the customer expected another auxiliary service to be included which was, which would have cost 15 rupees per transaction. So it's a bank card basically. So an ATM transaction, right. And they said my bank card gives me three free in the month. Right. So we, the next day we went back and said okay, it's 100 rupees a month and we include three free ATM withdrawals. And people were fine with it, right. Whether they used it or not. But you know, and then the next day we went and we sold to another customer and we doubled the price to 200 rupees per month and they were fine with it. Then the next day we tried 400 rupees and we literally went, every customer we just double the price and see where the breaking point is. Right? And in 400 people said that's too expensive. And then we said okay, you know, the CAC is so high, how the heck are we going to recover? And so we said what if we did a 10 card minimum or a 10 user minimum. Right. Which flew. And so that's how the price was discovered at you know, 2,000 rupees with the 10 user minimum. And it made the you know, sort of the CAC like you know, recovery period six to nine months and made it a very attractive proposition.

This is a great point. Discovery can result in price, can result in changing how you go to market and so on because the price you discover might be different than what you assume up front. So I think that's another reason why it's important to, you know, understand early on though that said, your portfolio economy, Mygate I think has an extremely, well, you know, monetizing much later, if I remember correctly, they had a free service for a very long period of time. So there are exceptions, amazing exceptions that prove the rule that you can actually not monitor.

I think in their case because they're sort of pretty much the only player in town and have a very sort of compelling, almost mandatory daily use case. Right. I mean, because you can't, you know, your Amazon package can't come home unless you approve. There was enough utility. Yes, yeah, yeah. And of course they've also been very thoughtful about when they start monetizing, you know, get the habit going and dependency. But yeah, you're right, those opportunities are kind of rare, I think for the most part where the customer has a lot of choice. Right. Because here the, the stickiness comes because 400 people in the community have got to change from, from their product to somebody else and there's always inertia there.

That's right, because what happens is the primary measure of PMF could instantly change if you start pricing. Unless you're a product like my gate. So that's why it's so important because you could say I have 80 retention. Oh yeah, that's for a free product. Let's see what price does to it. Right. Price could reduce it to 20% and then all your assumptions fly out of the window.

So yeah, look, I think this is a topic in itself. I know Sheepathi has written, my partner has written a blog on the high price of mispricing your product because especially in a category creating company, sometimes you price it solo that you just reduce the tam. Right. And you reduce the attractiveness. Although the value to the customer might be extraordinary. Right, so we'll switch gears a little bit. Go. Cool. It's just cognizant of the time here. There's so much happening on the AI front these days across the board and I broadly have been thinking about it as, you know, three things. One is obviously the platform plays, which they're very well documented and talked about. Would still love to get your quick views on those. But in terms of opportunities for startups and VCs, one is sort of what I call the scaffolding around the business. Things like customer service, automation, things like sales, marketing and things like that. And then the second is products which are inherently built on AI, some which are just enabled because of all the advances in AI and others where AI can make them much better business flows and things like that. Would love to get your take on how you view this landscape in the first place. And the opportunity space certainly for startups and which are the ones that you see as opportunities especially for India based companies that might be targeting US based customers. Exactly.

Yeah. I think it's a, you know, again it's a topic you could spend hours on but like you said, I think at the highest level there's infrastructure and applications with the middleware layer in the middle. I think the infrastructure piece is both chips, Nvidia of course, and also foundational models. I think the challenge with foundation models in general for any startup, let alone India or the US or anywhere in the world, is that the capital intensity, the data intensity and the compute intensity needed for these things is exponentially scaling. I think the estimate is that now you can, with GPT 3.5 like models you can now $10 million you can train. GPT 4 model will cost $100 million to train. GPT 5 is probably going to cost north of a billion dollars to train. And so basically, and then the thing is it won't change because it's an arms race. If you are a hyperscaler, a Google, a Facebook, an Amazon, etc. You're not going to want to lose this race because you don't. And even Apple, you don't want to be beholden to someone else's platform. So they are not investing. So if you're a startup, you're basically competing with the largest companies in the world who have made this a strategic priority and who have said they're willing to overinvest versus underinvest at the risk because they don't want risk missing these platforms. So you're competing with the four largest and most ferocious competitors and most capital intense and the most profitable companies in the world. So it's almost impossible unless you're Sam Alpin and OpenAI maybe, maybe, maybe anthropic. So I think there's maybe five or six companies in the world who are going to be able to invest in this for the next several years because it's not stopping. That's the thing. If you stop, you're done. You need continuous pool of capital. That's why I think most of Sam's time is almost certainly spent in raising capital at this point more than anything else. And you go to just like venture funds, became really large and started going to the Middle east and so on. You hear news of him going to Saudi Arabia and so on. And you, you probably have to do the same thing. You got to now get $7 billion, next $50 billion, who knows what. So I think that's why in general, it's not a fruitful endeavor for startups. I think chip layer, there are some competitors, we'll see how it goes. As you know, Nvidia has built a very compelling software platform also which is part of the lock in not just the hardware, but I think that generally is a harder nut to crack. So I think the opportunities are in the middleware layer and in the application layer. And in application layer there are two kinds of applications I think of. So in the middleware layer there's a bunch of tooling. I think the challenge I've seen with these tooling companies so far, I'm an investor in a few, is that adoption, especially amongst enterprises is a little slower than you're expecting. Why? Because enterprises don't really know what they want to build yet. So tooling is too early in many cases to test and so and in many cases even if they're adopting these tools, they're not paying much and so. And they're just using. Maybe the one interesting kind of tool you have is an orchestration layer that allows enterprises to switch between different models. But the reality is as models consolidate, people are seeing, well, I don't need 10 models, I have two or three models I need to use. So there's some stuff there. But many of the other tooling I think is going to be built into the foundation platforms themselves or the cloud providers. And I think the margins in tooling is going to be squeezed. There is some stuff I'm excited about around safety of models, around observability of models and so on that will need to be done by a third party because you can't trust the models to police themselves. So I think there's a system of tools around that and I think there's a good middleware around that. I mean invest in a company called Petronas that's doing some good stuff there. There's other companies there. But I think the app layer is the most fertile of all. Just like you mentioned, I think there are, there are various ways you can categorize apps. One of the best ways I think of is functional horizontal apps and vertical apps, industry specific apps. So there are apps that like if you look at accounting, AI, app accounting, AI applies across all industries in theory, but it's more of it takes a function like accounting and basically automates it as much as possible. But then you take an app for the construction industry or the auto industry that those are vertical apps. So I think those are two kinds of apps. Even within that I think the most Interesting opportunities are within the vertical ecosystem. Why? Because in functional apps there have been SaaS. There are SaaS companies that have automated over the last 10, 20 years. Many of these functions not with AI but with general like you know, just heuristic based software. There's Salesforce in sales and marketing, there's HubSpot, there's, there's like 20 companies, there's outreach, there's gong, et cetera in sales and marketing alone and each vertical quickbooks in accounting. Right. And these companies have customers, they have data and thanks to the foundational investments they've been able to easily spin up AI based functionality within their products or embed AI functionality. Now they're not as good as AI native products, but they're good enough, they're good enough that their customers are placated for a little bit while these guys are working on AI native products. So in general I think it is going to be fairly challenging for startups to break out. There's going to be exceptions obviously that prove the rule but it's going to be hard for startups to break out in functional AI company in functional AI categories such as sales and marketing, such as accounting, such as finance, unless there is no incumbent in them. I think anytime you have an incumbent that is between 5 and 10 years old, these companies are fairly innovative and fast moving. For example, I think in customer service AI which was a very fertile part, very fertile domain for AI in fact it's probably the one area where companies, customers are using it fairly actively to displace human agents or augment human agents. There's at least 20 companies. I think one of the YC batches had like eight companies and I think there's easily, and they're all focused on different things. But I know of at least two companies that were started five to seven years ago that within the last one and a half years they said okay, you know what, our old school customer service automation business is going to be disrupted. Let's just migrate all our customers, just build an AI customer service agent and migrate all our customers. And they are at almost 10x the scale of any of these startups in one and a half years because they already were at significant scale, close to 50 to 100 million and now they are at in, you know, in the double digit millions very, very quickly as a migrating car. And that's what you're going to see. I think there's a lot of startups that are raising rounds from great VCs but the reality is their scale is dwarfed by the scale of the incumbents, even if they move just 10% in the first year, that's more revenue than most of these startups are doing. And these are good products that they're launching these incumbents because it's not that hard. The foundation models have made it not that hard. And now coding agents, all of these together. So now all of this serves to say that I think the opportunity is in verticals and I think within verticals and in something that as you know, is called service. As a software is the other opportunity which in many cases deals with verticals where you take an existing service, a business process or service and you completely verticalize it. And so you basically make it an API endpoint where you can basically take almost an IT consulting service, say actually a margin consulting service. A great example McKinsey offers. You tell McKinsey I want, I'm going to pay a million dollars of McKinsey, Bain, DCG. You, you literally there are companies that are doing McKinsey as a service where you basically ping that software and it gives you a amazing McKinsey style report on something with a bunch of inputs that you give it. And so that I think that has the opportunity to take, I believe there's $2.3 trillion that are being spent on various IT and business services. So in theory you can take all of those services and make them as software endpoints, very very customized for that service. And I think Indian companies in particular are very, very are in a pole position because I think everything that Infosys offers or TCS or wipro you could do with AI without humans or at least augment them. And similarly everything that McKinsey offers, as you know, many of these large managing consulting firms have their teams in India and I think there's a lot of knowledge there. So I would be excited to see ex McKinsey folks in India leave and offer McKinsey as a service offer. Goldman Sachs analyst reports as a service offer, you know, Wipro or Infosys or TCS as a service, all those enterprise IT integration as a service. So I think software as a service is a huge option. The other one is what is interesting to me is I've been now meeting.

Services as a software.

To me, service services, software. I've been now meeting companies in India that have gone after US and European verticals and there's a company I met called Spine S P Y N E really interesting company. All their customers are auto dealerships in the US and Europe and they basically built merchandising software for these Folks and they have basically using a sales team in India been able to sell this pretty cool merchandising software and they are now have hundreds of dealerships in, in, in, in basically, no, not nothing in India. All their businesses in the US and now they're building a full suite of products. They race around, et cetera, et cetera. I was like, wow. I met, I met Sanjay the CEO and I'm super impressed with, with what he's built and what the team has built completely out of India. So I suspect, because speed of iteration is name too, right? It's a. Exactly. It's a great name. And so I think they're going to be. And I've met more and more founders who are taking, who are basically taking verticals in the US or in the west and going after it with just like Zoho and FreshBooks did with horizontals with functions sitting in India. I think you're going to see the next generation of vertical AI companies, many of them. I think Indian companies have as much a right to win as any company anywhere because they can build software fast. You need to build a bunch of AI agents. They have a lot of data and I think that, and I think the other thing Indian companies in general have a right to win I think is in fundamental needs in the Maslow's hierarchy, things like health, education, etc. Because you have a lot of interesting data and operations that you can train your business on in the Indian context and you can really bulletproof your product and your operations operating in India and then you can take those learnings and bring them to the rest of world in a much more cost efficient and operationally robust way. One of them is our portfolio company, Dozy, which as both of us know is building this AI enabled healthcare platform that is super low price but is incredible. It transforms any bed into an ICU bed and basically lowers the need for staffing, for dedicated staffing and nurses. The nursing is a chronic shortage everywhere in the world, including in India also. But I think definitely in the west in the US So they allow one tenth or one fifth of the number of nurses to be able to monitor a set of beds using AI, using signals, using just this is technology that they've proven for a few years in India and now it's used by all the major Indian hospitals and now they're seeing strong success here. And I think that model taking the same human needs, right? I mean the human, human beings have the same need for medical care whether it's in India or the US and now you have technology that is making it better. I think similarly, radiology, all of these things. Training data is so much more available in India compared to anywhere else because you have so many more lives and so many more people. So you can train your models much better in the Indian context and then you can easily leverage those things in other countries.

Yeah, actually it's interesting that I was chatting with Mudit and Gaurav the other day about the US market opportunity for Dozy and they almost have to dumb down the product, right? Because they have the product that they've got in India that is working here in the hospital. You know, people are running a very sort of strong operation on Dozy. But in, in the U.S. you know, now that they are FDA 510K approved and, and they're rolling it out there, I think there are subcomponents of the product that are, you know, extraordinarily more valuable to, to the users and they're almost building a simplified version or lighter version of the product because the market is ready for it, has been looking for and has already been using sort of not so elegant solutions. And so this is a very simple, you know, to the customer it's just okay, it's a better version of what I already know I need. Whereas here is a lot more in evangelization and stuff that needs to be done. So but you're right, I mean the benefit of training this and especially, you know, in the case of, you know, clinical grade products, right, it's, it makes a big difference. And I think we're seeing another young startup of ours called Zoo AI that's doing, you know, this IB curriculum like a self study buddy for, for students. You know, there are enough kids in India that are also studying the IB curriculum and now that it's trained here, you know, they're able to go to other markets. It turns out I was surprised that Turkey is one of the large markets of IB curriculum and of course the rest of the world, the US included is also huge. So they actually got their sort of their second batch of pilot users. You know, there's a lot of downloads from Turkey and a lot of usage and fine tuning of the product. So yeah, I think it's, it's a super exciting time. And on the vertical SaaS we've also had companies that, you know, they're working on, you know, good, you know, I would say operating systems for mid sized businesses, verticalized solutions. Like we have a company called Bookie that's doing, you know, the, how should I say the Studios for, you know, you know, small gyms. And those types of studios are small to mid size. And the AI innovations that they've layered into the product now have suddenly given them explosive growth. Right. So I think the customers also are looking for that kind of an edge. In fact, one of the thoughts we've also had is bringing in the services and software element because with the human in the loop in India, you can actually do a lot of like for example, the front office manager is the hardest role to hire and train and retain. Right. Well, that can be done, you know, with a combination of software and trained human in the loop. So I think some of these paradigms are going to evolve and I think it's going to create very interesting new opportunities.

And what is crazy is that many of the current companies today already are using their gross margins, actually 40% or 50% because they have humans in the loop. And so almost everyone using humans in loop too, get training data, use reinforcement, you know, learning, feedback and so on. And I think Indian companies can do it much more cost effectively and at scale compared to any other company.

Yep, yep. So having said that, there's also the, the risk for the larger Indian companies. Right. If, if so much of the software side of it, the development side of it, I mean, I guess there are two schools of thought. One is saying, hey, you know, there isn't a whole lot of very senior talent in India and the copilot approach might actually level the playing field and you might be able to get a lot more value out of the not so experienced talent. And then there's the other side saying, well, will you need this talent at all? So it'd be interesting to get your thoughts on, is there always going to be a continuous demand for talent? And with all of the stuff happening in AI, is there a future for software developers?

I believe so. I believe what you said. The first case, the optimistic case, what I believe in, in that the number one thing that AI does is narrow the skill divide, where it makes someone who's a junior developer equivalent to mid level developer and a middle developer into senior developers. So the skill premium starts going away. The notion of a 10x engineer or 100x engineer who's paid 10x the compensation of, you know, of a 1x engineer, I think you're going to start seeing that, seeing that shrink. And I think that's where India has a lot of, you know, 2x 5x, maybe fewer 10x engineers, but I think those 2x and 5x engineers can become 5x and 10x engineers using AI too. So I think because, because I think you still need engineers to build software, but you may not need like super expensive engineers. You might need just reasonably talented analytical engineers whose skills are augmented by AI. So I very much believe that engineer, that AI will augment engineers and improve their skills and reduce the skill premium, which very much benefits India because India, you know, and I think that, that I think you're going to see happen versus just replacing engineers en masse.

Wonderful. So let's switch gears a bit. Gokul, I know you've been a very prolific angel investor over the years and obviously love having you on board in a meaningful way at Dozy. And your your insights and inputs have already been very valuable to the team. But you're also transitioning now, or transition now into becoming more of a venture capitalist. Right. What do you see, as you know, similarities and differences from where you're largely investing your own money and can make a decision and look in the mirror and say what was I thinking? To when it's when to capital and there's an expectation and responsibility that goes with it. You know, it might Orban's like a hybrid of, you know, both yours and a fund that's raised with third party capital. How are you seeing the similarities and differences and any early insights into your new journey?

Dear listeners, thank you for listening to this episode of the Podcast. Subscribe now on your favorite podcast app for free and you'll be the first one to know when new episodes are available. Just search for Prime Venture Partners Podcast in Apple Podcast, Spotify, Castbox, or however you get your podcasts. Then hit subscribe. And if you have enjoyed the show.

We would be really grateful if you.

Leave us a review on Apple Podcast. To review, read the full transcript, find the link in the Show Notes.