Full Transcript

Dhruv Sharma: The Offline Network is streaming live. This, we're sitting here with Aman Goel who's the founder of a company called Grey Labs AI. It's a voice AI company with a focus on on the BFSI industry. And let's welcome Aman to the show.

Aman Goel - Co-founder & CEO, Greylabs AI: Thank you Dhruv. Thank you Utsav. Pleasure to be here. Welcome.

Utsav Somani: So why don't you explain to our listeners what Grey Labs does.

Aman Goel - Co-founder & CEO, Greylabs AI: Sure. So we are trying to build voice AI agents. Essentially, the idea is to automate the call center experience in a financial institution. So I'm sure all of us have faced a situation where you call up a bank, you explain your issue, you explain your problem, and for some reason, the call drops in between. And now you call up again, and you have to start the whole thing, right? That happens because there is no institutional memory in the back end. And the call gets transferred from one telecaller to the other whoever is available. AI can always be available. It can carry the memory of your last conversation, and it can make the conversation most very highly personalized. So that's what we are building, trying to automate the entire call center experience, be it telesales calls, be it customer support calls, be it, you know, collections call, welcome call, all types of calls of financial institutions using AI, which will talk to you in your language, explain in simple terms, it will remember your past conversation and make the whole experience much more personalized. For the enterprise, it leads to lower costs compared to humans and more consistency in service. For the end consumer, it is a way better experience. It saves you time, and you don't have to, you know, spend hours and hours on waiting queues in the call center.

Dhruv Sharma: Aman, let's take a step back.

Utsav Somani: I just want to clarify or get clarity on, is BFSI the industry where most number of voice calls happen?

Aman Goel - Co-founder & CEO, Greylabs AI: Yes, there are other industries also, there's real estate, you know, we all get these calls from, you know, some real estate companies, etc. But BFSI as a whole, I feel that finance is quite close to my heart. It's a large industry. It has heavy spending on the technology and banks are, if you look at, you know, the indexes, indices in India, Sensex, Nifty, etc., financial institutions are at the top, which indicates the volume scale impact. So, there are companies who are building horizontal platforms, but we chose financial institution for simple reason that the impact is very high. Plus, we are in Mumbai, you know, clients are closer to us, the proximity reasons.

Dhruv Sharma: That was in fact going to be my question, Aman, which is, I mean, there's a lot happening in voice, but why does this industry deserve its own voice company? But what makes it, what makes, what are some of the specific requirements your customers, your clients have that maybe other industries don't really have to think about?

Aman Goel - Co-founder & CEO, Greylabs AI: Sure. I think the biggest requirement is, you know, that many other industries are not regulated, e-commerce, real estate, etc. They don't have a regulator sitting on their heads. Of course, there is a regulator, but the regulation is much lighter compared to how rigorous it is in financial services. So, let's take an example of a bank. So, first of all, regulation enforces data localization within India. So, these call recordings of the AI model with the customer, like AI talking to the customer, this cannot go outside Indian boundaries as per regulation, if the data includes financial sensitive information. This will not apply in case of e-commerce or hotel booking or, you know, real estate call, etc. Number one. So, data security as a whole is a very big and important component in any software which is sold to financial services industry compared to other industries. Second, on our telesales call, right? We all, with financial services, let's say telesales, you get calls for personal loan, credit card, etc. You cannot miss sell any product. That's highly regulated in the financial services industry compared to other industries. Of course, it's a large scale problem and these things keep on happening even today. But this solves a big problem.

Utsav Somani: RBI came out with something very recently, right? RBI came out with, I think, something last week or maybe, I don't know, this weekend where they said that miss-selling, I mean, they put out a strict notice.

Aman Goel - Co-founder & CEO, Greylabs AI: So, someone 90 year old, 90 years of age person was sold an insurance policy, a long-term insurance policy by a bank, which was completely, you know, a big example of miss-selling. Because at that age, you don't need a term insurance policy, etc. I mean, you can argue you need it and all, but the policy sold to that person was completely in the wrong spirit. So, that's why RBI came up with such a rule. So, this usually does not hold again in e-commerce and real estate, etc., where money matters are not as deeply involved as in the financial services industry. Third thing on, you know, customer service side, the regulators are keeping a strict eye on what kind of personal data the call center agent is asking from you. CVV, OTP, etc. A banking call center agent cannot ask from you. The regulator keeps a strict eye on that. Fourth is that every call in our financial services call center is tagged as query, request, or a complaint. These complaints, so complaint, you know, we all can complain about a lot of things, but in the definition of financial services, the complaints are tagged in a certain manner. Like you try to, you know, withdraw money from an ATM machine, the account gets debited, but the cash doesn't come out. This is a good example of complaint. So, regulators want the complaint count every single month to be reported in a precise manner. You cannot over-report, you cannot under-report. Both are problems. So, these kind of smaller nuances, they don't exist outside financial services or they exist in a much lightweight manner compared to a banking or a financial institution. In stockbroking, you know, your application goes down. It's a big problem. Customers can incur financial loss. Indigo Airlines went down. This was the first time, I think, regulators stepped in and did something, etc. But flights keep on getting cancelled and nothing happens. Nobody will give you any voucher or anything you face inconvenience as a customer. Nobody cares. But in BFSI domain, banking, financial services, insurance, that's not the case. Even for small matters, the regulators are very strict.

Utsav Somani: And talking about regulated industries, and I refer to this podcast that Aryo, who's the founder of Anthropic, did with Rakesh. I think one point that I read about the podcast was that what is limiting them? I think the speed of AI diffusion is limiting them to some extent. Like, they can put 5 trillion tomorrow and have the best country of super geniuses. But eventually, people will have to adopt them. And corporates, individuals are much faster to adopt. Corporates are much slower. And I'm guessing banks and regulators and this BFSI industry will be the slowest of all. So, what does a successful pilot look like for you? How does that journey? Like, how do you reach out to, say, somebody like your client, Motilal Oswal? And what does that process look like for you? And anyone who's selling to a BFSI, any AI-based solution that's being sold to them?

Aman Goel - Co-founder & CEO, Greylabs AI: See, it's a long journey. Usually, the sales cycle lasts from three to four quarters. And then there is a delivery cycle, depending on the scale and size of the organization, right? Some big banks I've been chasing for maybe a decade now, or since 2017, college days in my previous company, still not cracked them and we keep on pursuing. So, the dynamics of a large enterprise are very different from a mid-market company or a small startup who is looking to buy a product. Because, see, when the company is slow, anything they buy, it only has an upside. But when you're a very large institution, there is a lot to lose if something goes wrong. People at the top, they need to care about it. It becomes a national news if something goes wrong, right? So, the first... Let me explain step by step how the process from starting till revenue comes in. First thing is establishing the touchpoint with the customer. So, usually, we try to find people in our network. That's the best way. So, some CXO of a bank or a financial institution who has some connect with us in some shape or form. Very simple example, someone who's an IIT Bombay alumni. So, I graduated from IIT Bombay. So, somebody who is also from my college, there is almost always a way to reach out to them, right? I'm a part of IIT Bombay WhatsApp groups where I can put up a LinkedIn profile of that person and say that, is there someone from their batch who can make an introduction to me? That's one of the easiest way. And also, even if you cold reach out to them, the success rate is usually higher. There's a soft corner that, okay, this is a younger version of me from my college and I should at least respond to them. So, lots of our customers, we have reached out via that part. Harshita, my co-founder, she's an IIT Kanpur alum. So, that compounds the network, exactly. We also have investors who are very deep in the financial services domain. They can open up and make warm introductions in their portfolio companies, right? We have Elevation Capital who are investors in many fintechs. They are investors in Paytm and a bunch of companies. Paytm is a big name I'm taking. Similarly, we have Z47 Matrix Partners. They are investors in a bunch of companies. So, they have been very, very helpful in opening doors. So, we consciously chose investors who are...

Dhruv Sharma: How about your angels? Do you want to give a shout out to some angels who are helpful?

Aman Goel - Co-founder & CEO, Greylabs AI: 110%. So, we have one of the angel investors, Nikhil Sami. He is one of the founding members of Yes Bank. And he was the country head of MasterCard. So, he is probably 20 years older than me. And he is also from IIM Ahmedabad. So, he has his IIM Ahmedabad alum network, Yes Bank alum network, and MasterCard, he was on the vendor side. You know, on Yes Bank, he's on the buying side as a customer of a service provider like us. As MasterCard, he's on the selling side. So, he has a lot of network. He has introduced us to several other prospects because they are peers to him. They are literally... They've worked together in some capacity. They've been a part of some forum. They've been a part of some event between the banks, etc. So, that's a starting point.

Dhruv Sharma: You know, I have to say, as Amon is speaking, I'm thinking the cap table is almost like a layer to CRM if you construct it well.

Utsav Somani: Yeah, I actually wrote an article on this. It's live on youridealround.com. Like, I think, I mean, it's a proper piece of equity that you get. I mean, land ownership that you can give in your company. And if you construct it well, some are like making noise on social media. Some are helping with introduction. Some are just brand, but don't respond to WhatsApp. So, it's a thing that you can construct very strategically. And I mean, you've built a great business. Of course, you're privileged enough to actually have that voice where you can pick and choose your sources of capital also, I'm guessing.

Aman Goel - Co-founder & CEO, Greylabs AI: Yeah, I think capital is one of the components of the value that angel investors can begin, bring in. Investors or angel investors both. But I think over and above capital, if they can do something. So, you know, we have Anil Gauteti, he's the founder of ScapiaCard. Very dear friend of mine. So, he comes from Flipkart background. He spent almost a decade at Flipkart. Almost all the fintechs will have some senior level person who is ex-Flipkart. You know, I was trying to approach one of the fast growing fintechs. The point of contact there was ex-Flipkart. I reached out to Anil. Anil is like, I was his reporting manager at Flipkart. It was an easy introduction. So, these kinds of things help a lot. So, that is the first point of establishing a touchpoint. We also have a SDR team, what we call as Sales Development Representative Team. So, they are a set of people who reach out to customers via WhatsApp, via email, via cold calls, etc. The conversion rate there, of course, is much lower the head rate. But then almost, you know, 50 to 60 meetings per month, because I'm telling as of January figure, they set up for us by cold reach outs. The success rate, again, I'm repeating is very low. But then it works for us. 60 meetings a month is a lot. And then, you know, you can hire more people and double down on this number, which we are very actively doing. So, the first meeting does not guarantee you revenue or business. But of course, it helps you get foot in the door. It gives you an opportunity to present your demo. So, that's when our sales team comes in. Sales team, along with usually I also go as a part of the first meeting, because these are introductions via my network, or some senior stakeholder on the customer side who wants to talk to the CEO. So, I personally spend a lot of time meeting people in the very first meeting just to set the stage, explain about who I am. And, you know, being a technical founder myself, it helps a lot. Because what is happening in the world of AI, people are unclear about what can be done, what cannot be done with AI. And salespeople are saying a yes to everything. They are overpromising. And later on, when you're not able to deliver that, it creates a situation. Customers hate that. But with my technical background, meeting the customer in person, discussing some idea, brainstorming it, it becomes less of a sales conversation. It becomes more of a technical conversation with someone who knows the technology very well, or, you know, who has experienced it. So, you know, I was talking to one of the banks. Their head of AI has saved my number as Aman Goyal Gen AI expert. She said, I will not save your number as Aman Goyal Greylabs AI. I have saved your number as Aman Goyal Gen AI expert. Several clients invite me to take sessions on AI for their teams where we don't sell our product. We talk about AI in general. We educate them what can be done, what cannot be done. We go beyond voice AI and explain them. It's not a sales meeting at all. But what ends up happening is that it builds a lot of trust with the customer, with me being involved in the sales process, being a technical person. And I have my sales co-founder, Shivam. He, of course, then drives the rest of the deal cycle, which means showing demos, you know, understanding the requirements of the client, helping them understand specific features in the product, the data security aspects. So we typically have seven stakeholders in our buying cycle. There is a business stakeholder, IT, information security, legal, compliance, procurement, and finance. And now with voice analytics and all, there is also usually a data analytics team getting involved. So there are total eight stakeholders in our sales process. So our sales team then takes care of that. We have roughly, I think, between sales and SDR team, we have about 20 people as of now. We are, you know, adding people every quarter, every month. So they will then meet the customer in person. They will explain them. They will show a demo of the product. They will give credentials access. They will try to do a proof of concept with the customer, et cetera, et cetera. Then, of course, you know, the procurement process where clients may want to invite multiple service providers, they may want to understand your differentiation, et cetera, those kinds of things kick in. Wherever the conversation gets stuck, I try to step in or I use my network wherever I mentioned the name of Nikhil, I mentioned the name of Anil, I mentioned Z47 Elevation. So wherever, of course, there is a complete roadblock, we try to take their help to understand. Sometimes customers are not comfortable explaining the whole situation, right? So these people, they help us understand what is going on. So depending on that, we prioritize our efforts and that's how it goes.

Utsav Somani: Awesome. And what's the state of the business right now? Give us a, I mean, paint a landscape picture of how many clients you're working with, what kind of clients, what's the scale in terms of revenue and stuff?

Aman Goel - Co-founder & CEO, Greylabs AI: So we are upwards of, so we have two products actually. One is the voice AI calling, which I talked about. We also have one product, which is voice analytics. What it can do is, so, you know, today, if you have a human call center of, let's say, 500 people, who is talking to customers across different use cases, this product can automatically audit those calls for quality and compliance purpose. So, you know, when you call up a bank, it says this call will be recorded for training and quality purpose. Today, what happens is there are human auditors who take a sample of calls of every telecaller and then they fill a quality sheet. Something like they listen to the call and then fill an Excel sheet which says, did the agent talk politely to the customer? Did the agent authenticate the customer before sharing any sensitive information? At the end of the call, did the agent, you know, ask for feedback? Did the agent ask for CVV, OTP, or any sensitive information which was not supposed to be asked? So all these things, they are done manually. With our software, we can automate 100% of this procedure. And also, you know, these calls are across different languages. Imagine you have some customers in some, you know, very simple example, you have some customers in Maharashtra. So you need to have an auditor who understands Marathi language because the call recording is in Marathi. So India is a diverse country. There are so many states, so many languages. You have to have auditors for each language. With AI, you can save that cost. You can scale up the audit to 100% of the calls. Besides auditing, we also give consumer insights. So why are the customers not purchasing? Why are they unhappy? Did they talk about any competitor on the call? Did they mention any pricing issues on the call? So that also gives you insights into how do you improve your product offering as a bank or as a financial institution. So between these two products, we have more than 50 BFSI clients. Some of the logos you can see on our website, but we have, you know, banks wise, we have RBL Bank, we have IDFC Bank, we have Federal Bank, we have DCB Bank, we have Axis Bank, we have Motilal Oswal, we have Grow, we have Slice, we have Credit Bee, Bank Bazaar, PhonePe, BharatPe, bunch of such customers. So that's on the customer side. We are roughly 65 people in the team, largely in Mumbai, about 50 people, four or five people in, I think, five people in Gurgaon and about 10 people in Bangalore. And we are gradually setting up our engineering and product center in Bangalore. And we are about two years old. Between seed round and series A round, we have raised about 100 crores in funding in about a year since starting.

Dhruv Sharma: Aman, you know, when people talk about voice, one thing we hear very often is if there's money involved, I want to talk to a real person on the other side. And we get that, right? So that's maybe the down case for speaking with an AI voice agent. But what's the other side? An AI agent, for instance, will never talk down to you. It will never patronize you. It can be more empathetic. So talk to us about that and maybe how voice can actually take us to a vision where regardless of where your customer is situated and what the scale of the banking relationship is, you can actually treat them really well.

Aman Goel - Co-founder & CEO, Greylabs AI: Sure. So let's split the conversation between the benefits to a bank and the benefits to an end consumer. Actually, many of them are aligned. The optics is just different, right? If I, as a customer, walk out frustrated after a call, the bank also loses because, you know, as a customer, if you're unhappy, you will talk about it on social media or you will go to some other bank. So we are actually aligned. So first of all, the language is one very important part where you can scale the call center across the country, across multiple languages, across those kinds of barriers. That's point number one. Point number two, AI will never be fraudulent or it will never try to miss sell you any product which has been programmed already in a certain manner. Right? So every time, by the way, let me tell you, there is about 40% churn, employee churn in most of the call centers today. I know companies where in six months, the entire team is replaced, 100% churn. So if you have a thousand people call center, you need to hire 400 people every year just to keep up the number and then hire more people because your company is growing 15, 20% year on year. Right? So the hiring of those people, you know, the training costs, the training procedure, the training costs, two to three months worth of training. And then, you know, the real estate, laptop, all the software access around it, people problems, right? Two people getting into a conflict, somebody misbehaving with a customer. All those things come as a part of that. With AI, you can train it once on a certain procedure. So let's say a simple AI, which can talk to you and pitch a credit card product. Very simple. So you have trained, you have 10 credit card products. You train your AI model on top of that and then it will call. So now you can scale up and scale down as and when required. It's Diwali season. You can scale it up overnight. You know, more people are buying credit or you're an insurance company where January, February, March is usually the best quarter year and tax saving season, et cetera. You can scale up your call centers very easily in no time without having to worry about the real estate costs and our training costs and a bunch of things. Right? So for the enterprise, that is good. For the customer, it is good in terms of reducing the wait times. Otherwise, you know, we all have to wait in the IVR long queue, et cetera. So these are a few things. There is also, you know, there is a area of call center, which many of us may not be aware, which is the collections call center. In banking, what happens is when you don't pay your credit card or loan EMI on time, credit card bill or EMI, they will call you to collect money. Collections agent calling, right? It's a large use case. You would be surprised to know how many people are not in a position to pay their bills. We all, I mean, I'm a heavy credit card user, but I use it for reward points and other benefits and not because I need that credit. So probably many of the listeners here may not be aware of this, but there are large collections call centers whose only job is to call up the customer and somehow recover that money. Now, sometimes customer may have just forgotten.

Utsav Somani: It's not that pretty, I believe. Like, I mean, borderline harassment, if not more.

Aman Goel - Co-founder & CEO, Greylabs AI: Exactly. So sometimes there is an intent issue on the customer side where customer doesn't want to pay. They have an intent of default. And sometimes it's a situational problem where, you know, you are traveling and you don't have OTP because you are an international travel. You don't receive the OTP or you just forgot to pay. So the former part, their customer has an intentional, you know, will of default. They want to default. There, these calls oftentimes go aggressive. The collections agents, they are paid incentive on recovering the money. So sometimes they will abuse. Sometimes they will misbehave. Sometimes they will try to threaten the customer. Now you want to recover the money as much as possible because this is why you lent the money. But the regulator says that you can't misbehave with the customer in any manner. How do you threaten the customer? And where is the boundary that you draw where a telecaller doesn't breach that boundary? There have been situations where telecallers threatened the customer so much that the customer committed suicide. Can you imagine? So for fintechs and for fintechs is a, it's RBI, of course, keeps a close check on fintechs because they somehow, you know, try to run fast and breach the boundary. So that is one thing where AI can help. And second, on the banks and financial institutions, large companies, it's the same, right? RBI keeps a close check, but you don't want any untoward incident like suicide to happen with an end consumer because the regulator is not there to regulate just to regulate the bank or the fintech. It's there to protect the consumer. It's a government body. It works for the people. So with AI, you can save all of these problems, all of these kind of harassments. AI will never abuse the customer. AI will never threaten them. You know, AI can be more persistent, which means it can be more patient. It can be more persistent. It can make you more number of calls. It can, you know, that way.

Utsav Somani: So if you were to look out five years from now, is the world AI assisted service delivery or do you think it's AI only service delivery?

Aman Goel - Co-founder & CEO, Greylabs AI: See, I think it will be because some of the scenarios you still require human intervention as the technology improves. Let's see where it goes. But I feel there will be a combination of 70, 80% of the work which will be done by the technology and 20% you will require humans for corner cases. See, I'll give you very lame examples. In financial services, there are a lot of back end systems where they don't have, you know, where there is no API available or a connector available to be connected with AI system. So lots of such situation where integration is not possible, where there is a basic technology limitation rather than an AI limitation. So those things will take time to automate.

Dhruv Sharma: If the latter comes true, so then it'll be all of our agents having this conversation on the offline network. I mean, the next time you're coming to the show, you should also bring your agent or just send your agent.

Aman Goel - Co-founder & CEO, Greylabs AI: Send your wife. I think he's one of the creators. He has created his AI clone and generating videos out of that. That's what I read somewhere.

Utsav Somani: Yeah, and I think he's very much possible. I think a lot of people are doing these AI influencers on Instagram as well now. But a couple of questions. So what does the tech stack look like? Who are the partners? It's India AI week. So all the big names are here. Who are the names that you work with?

Aman Goel - Co-founder & CEO, Greylabs AI: Sure. See, we unfortunately cannot work with all these US partners, etc. because of data localization rules that I talked about. So our entire stack is built from ground up. We use a lot of open source tools. So we have Metalama 3.3, 7 DB, which has been fine tuned for BFSI. We have, you know, OpenAI Whisper, which is a speech recognition model, which has been fine tuned again on financial services data. So we use a lot of these open source tools. But yes, on the cloud side, we work with NASA AI, which is our cloud infra.

Dhruv Sharma: And used today, yes.

Aman Goel - Co-founder & CEO, Greylabs AI: Yeah, exactly. So we work very actively with them. We have AWS, we have Azure, we have Google. So all these companies, they offer some cloud credits. They have different benefits of scale, cost, etc, etc. So we work with a bunch of them.

Utsav Somani: And Sarvam?

Aman Goel - Co-founder & CEO, Greylabs AI: Yes, we integrate with Sarvam in the backend. So what happens is, I'll tell you many times you approach a customer and they have already onboarded some AI infra partner. So see, I'll tell you one. One is the GPU infra partner, which is NASA, E2E, Google, AWS, Azure. On top of that, there are AI infra partners, which is, you know, Sarvam, your 11 Labs, Cartesia, DeepGram, etc. They provide AI infra components. On top of that is the application layer, which is us, where we build the workflows, we have the dashboard, analytics, etc. So underlying, we can switch between the partners or between our own offerings, depending on scenario, depending on case, or depending on what the customer wants. The customer says we already have Azure cloud. And of course, I will have to deploy it in their Azure cloud rather than using NASA.

Utsav Somani: I think we're almost at the end of time, Dhruv. Any final quick question for Aman?

Dhruv Sharma: Oh no, I think we've had a good run with Aman today. Thank you again, Aman, for coming and sharing. We're going to maybe need a part two to this in a few months.

Utsav Somani: And we will 100% congrats on everything and including the recent 85 crore as well. So all the very best.

Aman Goel - Co-founder & CEO, Greylabs AI: Thank you so much, Utsav and Dhruv. Really enjoyed the conversation and looking forward to more such sessions.

Utsav Somani: Likewise. All right, listeners, let's welcome our next and the final guest for today. We have Sohas from Localhost. Welcome to the show.

Suhas Sumukh - Co-founder & COO, LocalHost: Hi, folks. Nice being here. Hi, Dhruv.

Utsav Somani: Man, you're super young. I believe you had something to do with university or a class or an exam today. That's why you couldn't join earlier, right?

Suhas Sumukh - Co-founder & COO, LocalHost: Yeah, exactly. I haven't dropped out. I still attend university. So that's why I was finishing up classes and then came to the last segment of the show.

Dhruv Sharma: Amazing. Let's take a moment. I think we have the youngest ever TUN guest like ever.

Utsav Somani: Yes, that's amazing. But tell us a little bit about your Localhost setup. Like you do Hacker Houses, you do a bunch of things. So I'm surprised you haven't dropped out yourself right now.

Utsav Somani: Not that I'm discouraging that, but please.

Suhas Sumukh - Co-founder & COO, LocalHost: Fair enough. So yeah, we started Localhost in 2023. We started off a Desort server for students and engineers, student engineers, developers to come meet people. We also started hosting hackathons along with them. And slowly realized, you know, when we started hosting these hackathons, a lot of folks came here for the money and not for anything to do with building products or like taking part in the process of building it. And that's when we realized, you know, we could actually fit a venture model here and actually fund people for their idea, rather than them competing for like a cash price pool or any sort of cash incentive. And that's where we started Localhost. It was mainly where we used to give a bunch of grants, investments, micro investments to like college dropouts, college students, and a bunch of 25 year olds, around 25 year old founders trying to build something. And then in 2025, we realized we were still online and we wanted to like focus on offline experiences. And that's where we started the Hacker House, which you all would have probably seen. We did that in Bangalore. It went super viral. It made international news. And we realized offline experiences can bring a lot more people. And also it's a good community engagement. And we copied the same model in Europe, in US and Japan. And slowly we started doing events on the side. So events became our talent pipeline, go-to-market engine for all the venture stuff we do at Localhost. And now starting 2026, we have closed all our Hacker Houses in US and India and are moving into a founder's lab. So right now, this is the new founder's space in Bangalore.

Utsav Somani: And you've raised Localhost Ventures as well, like which is a venture fund also.

Suhas Sumukh - Co-founder & COO, LocalHost: It's a separate pillar from the media arm.

Utsav Somani: Okay. And so you're investing in people who build at these Hacker House events?

Suhas Sumukh - Co-founder & COO, LocalHost: We invite these builders to come to the house and build here. So we have a thing called Community Fund. We give approximately $5,000 to $10,000 grant per person to build from like IdeaZero or Queen Corporation and post that if anything succeeds or go big, we will do a follow-up round and connect you with a bunch of investors. So that's how we have structured it.

Dhruv Sharma: So I also want to ask you about these founder labs. You're sitting in one right now. Yeah. And so a lot of people try doing this, but almost never get it right. What have your learnings been from having run the Hacker Houses and what are you trying with this one so it works?

Suhas Sumukh - Co-founder & COO, LocalHost: We have, we follow this thing called multidisciplinary people. So we not only bring core hardware founders or any founders here, but you also have people who write research papers, who are doing research in AI, who are policymakers, who work with the governments. Of course, SaaS builders, FinTech founders, general founders, and core hardware and deep tech founders. So they're all in the same floor. And this brings a very multidisciplinary learning with all of them. And given our incentive of cross-border...

Dhruv Sharma: Is there like an application and an evaluation process?

Suhas Sumukh - Co-founder & COO, LocalHost: We do an application and we also do referrals. So we go by referrals and applications. And majorly we come, majorly we get the founders to come back from our events, the demo days we all host. So a bunch of pool where we kind of invite these founders to come be part of the cohort or the program.

Dhruv Sharma: Can you give us a sense of the coolest things people are working on right now? Yeah.

Suhas Sumukh - Co-founder & COO, LocalHost: Yeah. So right now we haven't started any, our previous batches had people working on hardware, such as AR glasses. A founder from Mumbai was building AR vision glasses. There's this 18-year-old kid working on active noise cancellation system. He patented this technology when he was 15. Then there's the 16-year-old kid who's working on cursor for hardware. There's this company, Prava Payments and Maya Research, which came out of house. So a couple of these, notably you would have seen on Twitter or would have heard of them.

Utsav Somani: And I mean, so you're basically a Gen Z, if I'm not mistaken.

Suhas Sumukh - Co-founder & COO, LocalHost: Yeah, I'm 2007.

Utsav Somani: But Gen Z is investing in other Gen Zs. Do you think you're at the cutting edge of technology just because you're younger and more closer to these hackers, I mean, hackers and builders and tinkerers of the future?

Suhas Sumukh - Co-founder & COO, LocalHost: More likely I would say we are much closer to the ecosystem being built. The young people who have been inspired by the dot com era, the young Steve Jobs and the young Elon Musk stories. We are like very close to those people here and that makes a very good community of people who are ambitious enough to go in and build something of their own.

Utsav Somani: So tell us like a couple of, I mean, if you were to pick like three winners from everything that you've done so far or the hacker houses or investments that you've done on people that you supported via the Media Lab, what are the three companies that you would like to highlight with names?

Suhas Sumukh - Co-founder & COO, LocalHost: One, Maya Research. There's a cool story of him. He finished his master's in New York University and he was planning to build this company, Maya Research. It's a voice model system, voice model for India, Indian voice agents. Similar to Sarvam, but they are like a different benchmark and they're working on multilingual voices. But yeah, he emailed me when he got to know about me through Twitter and he emailed me that he's finishing his master's in New York University and he's coming to India and he wants to build. And so he was part of the hacker house there in Bangalore like last summer and he built this entirely pre-incorporation, everything done from the house, launched the product and in a month or two, I think he got like over 6,000 downloads on Hugging Face and I think South Park Commons also invested in him after that. And today it's like the world's top 20 model on the benchmarks. And South Park Commons One more company is...

Utsav Somani: They run like a minus one to zero community as well. We'll talk about them once you complete the other two.

Suhas Sumukh - Co-founder & COO, LocalHost: Yeah. So one more is Prava. Prava is building agentic systems, AI agentic payment layer. So basically like charity can actually go and buy stuff for you rather than just giving you product demos. They're working with like large merchants like Visa and Mastercard and are also going very big in the US market. And they recently got funded by Nikhil Kamath to the new WT Fund cohort. And the last one I would say is Markov. Markov is the RL environment agent. It was part of the Versailles Accelerator recently. Now they're like pivoting a bit, but they had a very solid on human data around robotics and human scaling mechanisms.

Utsav Somani: Amazing. Thank you for sharing that. I'll have one ask. Can you connect me and the team with these founders just so that we can have them on the show as well? And we'd love to spotlight anything that you want to invite to the show just so that we can share more and learn from them about these new technologies that they're building. I think just fascinating to hear young builders on the show.

Suhas Sumukh - Co-founder & COO, LocalHost: Yeah, for sure, for sure. 100% I will connect you guys.

Utsav Somani: Yeah, Dhruv, over to you.

Dhruv Sharma: Oh, yes. Suhas, I don't mean to be an ageist or anything, but these days when old people get together, they're forever talking about AI is going to take jobs away and entry level jobs are going to be a problem and this and that. But with folks like you, right, you're super young. You said you're 2007 born?

Suhas Sumukh - Co-founder & COO, LocalHost: Yeah, I'm 2007 born.

Dhruv Sharma: Great. So between you and your friends, what do you, when you guys talk about AI, what are you talking about? Does it make you nervous in any sense, like, you know, coming into the real big world? And I mean, yeah, let's just start there and then maybe we'll have a few other questions for you as well.

Utsav Somani: Yeah, because that article about something big is happening. I think you must have seen that on X as well. And then Dhruv and I discussed it, I think, on our last show, where I think they're trying to basically paint a picture where I ask everyone to just wake up around you and like use tools and just, I mean, upskill themselves.

Dhruv Sharma: Like, I think it's hard for anyone who started their career, you know, like 10 or 15, 20 years ago to put their themselves in the shoes of people who are still on campuses and who are, as they say, AI native. I mean, like, you know, you're still on campus and this thing is new and enhanced. So anyway, tell us how you guys think about it to us.

Suhas Sumukh - Co-founder & COO, LocalHost: Yeah, first of all, we are like very excited about how things are turning. Like past six months back, the AI world is like rapidly moving ahead. I would give like an example with film festivals, which we are doing at Localhost. The last one in Mumbai was like the first one we hosted.

Utsav Somani: We actually had both Localhost and Sanket on the show on Friday.

Suhas Sumukh - Co-founder & COO, LocalHost: Yeah, I just saw that and I was like giving you a good quick context. So in Mumbai, when we hosted this in October, the films were like pretty decently fine. But now when we see the same thing on the updated platforms and new mods coming out, it's a whole lot enhanced. And the world is moving really, really fast. In like next six months, we don't know what will happen. And we are like really excited about what's happening. I'm not like talking in perspective of jobs or like what's happening in the job market. But given the innovation and opportunity to do so many things, you know, the cost of building anything now, a company, a tool or anything is like really low. And it's also beginner friendly. And I see a lot of people trying to innovate a lot of new technology out of the AI, which is like really growing fast. So I see it's a very interesting trend to follow. And also like build, this is like really a good time to build something in AI space.

Dhruv Sharma: The AI has effectively like collapsed those learning curves that you earlier on had to like straddle for multiple years. But do you think sometimes that gives people an illusion of understanding as well that they think that they know more than they actually do at the moment? Do you think that happens?

Suhas Sumukh - Co-founder & COO, LocalHost: Sometimes if a person does not understand the basics, I would say, let's say a person who's just starting out to code, he might not know the basic fundamentals of what a node files are all of these. But a person who already has learned all of these basic fundamentals, I don't see an issue where they can actually enhance all of these and build really great things out of AI. And they are like an assistive tool, you know, just like second human brain. You have the idea and they execute it and you can see where things go wrong. And you could have seen already how far away it did over Twitter and the internet. And that's like a real prime example of what people can do.

Utsav Somani: And what is your tech stack, personally? Like, I mean, what are Gen Z's using these days? Are there tools that me and Dhruv can also start using? Apart from...

Dhruv Sharma: If you ever train us, what are the cool kids up to, Suhas?

Utsav Somani: Yeah. What are the cool kids up to these days?

Suhas Sumukh - Co-founder & COO, LocalHost: I think PlotCode. PlotCode is what everyone's using for web coding. There's also CodeX.

Utsav Somani: What about Cursor and all the others which were there before that?

Suhas Sumukh - Co-founder & COO, LocalHost: I've personally never used Cursor. I'm admitting it first time on live, but I've never used Cursor. I've only used PlotCode just because it's really good in narrative and reasoning, understanding of the code syntaxes. So I see a lot of people using PlotCode. And there's also a ton of these new AI co-pilots running on the internet, like Tracer, Cursor. There's a few YC companies now building the same thing. But honestly, I would say PlotCode has done a really great job understanding code and building stuff. And there's also CodeX from OpenAI and all the other tools like CloudBot and all of these are what people are using these days to build and ship web-coded platforms and tools.

Utsav Somani: And you've done Hacker Houses across the world now. What have you noticed differently? I mean, builders in different countries doing different things. What are you excited about? Certain characteristics shine in Japan, certain in India. What are the key highlights and trends?

Suhas Sumukh - Co-founder & COO, LocalHost: There are different type of geography we target and based on what sector they are doing. Like, for example, Japan, we have a lot of people building in consumer apps, a lot of folks building in like FinTech. And in France, we have people building in robotics and hardware. And in the US, it's a bunch of AI, SAS, here and there, a little bit of deep tech. And in India, it's also similar to the US. It's very multidisciplinary. And a lot of folks do mix and match hardware and AI here. But in France, I would say the biggest difference what we see in India is the talent density. People have the appetite to leave their jobs and work on their startups. So a lot of people living in our Hacker Houses are like ex-OpenAI, ex-Google, ex-SpaceX engineers who wanted to start entrepreneurship or do something in those. But in India, we don't have those. These are like really... Are they like locals? Are they what are called digital nomads? No, there are some locals and some people who moved from the States to France, which is really interesting in terms where people talk about EU's regulation in terms of startups. But there are a few people actually going to Paris. And Paris is also considered as AI capital for EU. So a lot of things are happening there. If you've probably heard of Station F, there's a lot of incubation of startups there. A bunch of cool companies coming out from Station F.

Utsav Somani: committed 30 million dollars to the cause as well. 30 million euros. Grand sum of 30 billion euros. 30 billion euros. But what are the cool AI companies from France? Why are they calling them the AI capital of Europe?

Suhas Sumukh - Co-founder & COO, LocalHost: Um, I don't dare... Mistral is from there. Yeah, Mistral is from there. I assume even like, the Lovable is also from Europe somewhere. I'm not sure if it's like France.

Utsav Somani: Lovable, I think is Sweden.

Suhas Sumukh - Co-founder & COO, LocalHost: Yeah, but I would say like there's a lot of like deep tech happening there. Like say hard robotics or like anything in between physical intelligence and AI. The physical hardware equipment is being built there. Very new market which we have tapped into. And we're like still learning a lot of things there. And we plan to set up a hardware space in this coming summer in Paris for the same cause of all these new builders coming and building hardware there.

Utsav Somani: Awesome, Swash. I think this was phenomenal. We've learned so much and you inspired us. Whenever you do your next outing in France and stuff, maybe we'll have you back on the show. Dial in from your Paris outlet.

Suhas Sumukh - Co-founder & COO, LocalHost: Sounds good. Yeah, guys.

Utsav Somani: Cheers. Thank you so much and all the very best. All right, listeners. Now that we've done our roster of guests today, we've got a couple of news items. It's the India AI week. So we'll, of course, be heavy on AI. OpenAI has acquired OpenClaw creator Peter Steinberger. I mean, it's an acqui-hire basically. Did you read into this, Dhruv?

Dhruv Sharma: Yeah, I mean, we covered it in the show as well, right? When we had Shrijan and Arun come talk about it. That was like 15 days ago. No one saw this is how it's going to end.

Utsav Somani: Man, like, it's insane. Like, I mean, this guy apparently failed 42 projects before this 43rd one, which is now, of course, become ClaudeBot and Moldbot. I mean, it went through a bunch of changes. And Mark Zuckerberg was also quoting him. I mean, apparently, the TechCrunch article says that Zuckerberg was reaching out.

Dhruv Sharma: Sliding into his DMs or whatever they say.

Utsav Somani: Sliding into his DMs. And Claude, or Anthropic, was sending him legal notices. So he was in touch with all the major labs. Only OpenAI and Sam Altman managed to get this deal done. But this might actually...

Dhruv Sharma: Like, Suhas and his friends would say, like, Anthropic got no chill.

Utsav Somani: But, dude, I think the crazy thing is that this might be the only solo... I mean, everyone was making comments about...

Dhruv Sharma: One person, one billion dollar company.

Utsav Somani: So this might act... I mean, we don't know the numbers yet because they've not disclosed it. And OpenClaw, the software, is going to be inside a foundation underneath OpenAI. But it's probably, I think, close to that number. Like, in terms of...

Dhruv Sharma: This is the first M&A between two foundations, right? Where a foundation has acquired another interest in another foundation, yeah.

Utsav Somani: But, I mean, now at least, I mean, there were technical challenges. Cisco reported that there's some data issues and stuff in this thing. So now, I think, with the resources that they have at OpenAI, I think all of that might get sorted out. And Manos, which was an AI software company acquired by Meta as well, they've released a very nice implementation. I actually was playing around with it on the weekend. Well, it's one click install for OpenClaw. Many others are doing it as well. But this one's super easy and simple. And even Kaman Bharti Mittal, I think, was experimenting and he ran up a bill of a crazy amount. Like, people typically report, like, I mean, going up to $5,000, $10,000 for token costs if they don't let the... don't control the bots properly and you let them run wild. So I think the cost will also be a concern that you have to meter in, basically, when you're using some of this.

Dhruv Sharma: People say that in AI right now, we're in the, whatever, $2 Uber era. Like, you know, back when they just launched, like, everyone's heavily subsidized. Another five, six years from now when this becomes very much an integral part of our workflows and they ramp up pricing, we'd have to pay a lot more than we do right now.

Utsav Somani: We'll have to. And I think, I mean, you sent me that link to that Dario podcast. I didn't listen to it because everyone on Twitter makes summaries of it. So he said that he'll have to predict, I mean, accurately predict demand because I think overspending on compute and you're dead or underspending and then you've lost out to competition. So I think that will be a very interesting dynamic for how this CapEx build happens for the world of AI. Did you listen to the podcast?

Dhruv Sharma: Uh, no. The one with Dwarkesh, is it? Dwarkesh, yeah. Yeah, me too. Or we'll just put it through Notebook LM and just get some sort of help for you.

Utsav Somani: Another headline coming. India has 100 million weekly active ChatGPT users. Sam Hoffman did an op-ed in one of the newspapers today. And they've opened an office in New Delhi and they've launched ChatGPT Go. I'm hoping that they do some more announcements across this week when I think OpenAI is speaking on Thursday. Sam Hoffman.

Dhruv Sharma: Intuitively, what do you think? Is ChatGPT like the dominant AI app in the country right now? Is this the one most people know and use? They have the mindshare.

Utsav Somani: And I think they raced away by giving that ChatGPT Go plan for a year away, right? I think that basically gave them a lot of redemption. And I mean, it became a verb. Like, I mean, you don't Uber. I mean, you don't ride hail it, right? You always Uber it. So ChatGPT is literally the first word that comes to mind whenever you're having a conversation with people who are not in our own bubble. So I think it's become that de facto tool.

Dhruv Sharma: And even this OpenPLOR acquisition is, I mean, of course, they have different businesses, their research lab at the end of the day, they have different businesses. But the OpenPLOR acquisition is just them doubling down on their sort of like consumer AI offerings. Yeah. Personalized agents for everyone.

Utsav Somani: So yeah, India AI Impact Summit. It's day one today. Yes. Big names. Like it's insane. I mean, we've covered this before, but literally Anthropic, Sundar Pichai, Mukesh Ambani, Nandan Ilkeni, Macron, UAE Crown Prince. I mean, Jensen pulled out over the weekend. People were not happy about that from NVIDIA.

Dhruv Sharma: Yeah, he managed. I mean, he isn't coming, but who knows, maybe sometime through the week, we'll take the offline network on the road and chat to our listeners.

Utsav Somani: All right. Blackstone backs NASA in a $600 million equity deal. And there's a covenant where they have to pay $600 million in debt as well, which will make this company a unicorn. And it's an infra company. Dhruv, you'll have to break it down for me.

Dhruv Sharma: NASA AI, right? In fact, Aman who just came on also, I mean, he's, theirs is one of many companies that works with NASA. Like the large private equity firms in the world have had their eyes on data centers for quite some time. So at the end of the day, part of that business is also a real estate business. And it has, I mean, you were just making a point about demand going through the roof. So it's only natural for capacity to also keep pace with that. And at our end, we've had, we obviously have, we were just talking about it on the show. We have data residency laws. And it's not just BFSI, by the way, with new privacy law coming into effect, everything's got to just stay here. So there is that. And then also you can't, I mean, you've got like, it's got to be close to its point of usage, right? There's latency concerns as well. So you've got that. And yeah, so this is one of the large domestic companies.

Utsav Somani: Let me throw some numbers. They've raised 50 million before and Blackstone estimates India currently has less than 60,000 GPUs deployed. But they want to scale that number 30X to 2 million GPUs in the coming years. NASA currently has 1200 GPUs and targeting 20,000 plus. So I think it's exciting because Blackstone building a global playbook or portfolio of these companies, data warehousing and AI.

Dhruv Sharma: I think the India AI mission, that cluster has somewhere between, somewhere close to either 34,000 or 38,000 GPUs. Okay.

Utsav Somani: So yeah, so approximately that I think tracks.

Dhruv Sharma: Oh, wow. And just privately, and so this will allow people to come train their models, fine tune them, deploy their workloads, everything, yeah.

Utsav Somani: No, solid to see this announcement coming this week. And there's another thing that was announced. India has approved 1.1 billion state-backed VC fund.

Dhruv Sharma: Oh, yes. This is like the 2.0. Fund of funds, yes.

Utsav Somani: Previous one deployed, I mean, quite a sizable sum across 145 funds, backing 1370 startups. And yeah, I mean, almost 2.8 billion deployed there as well.

Dhruv Sharma: Yeah, I mean, that fund of funds is anchored, most first-time fund managers is as their LP. And yeah, now there's a second edition. And this isn't, this is, there's also the RDI fund.

Utsav Somani: Yeah, I think we've had 49,000 startups registered just last year. So I think the story is just getting started. And it'll be exciting to see all these infrastructure announcements in AI that are going to come out this week. I think peak 15th also backed a company called C2I Semiconductors. And Bihar, I think yesterday, state of Bihar also announced that they're giving extreme benefits to companies setting up in the semiconductors value chain as well. I think that paper came out yesterday.

Dhruv Sharma: There are incentives. In fact, data centers reminds me, we'd cover that story as well, where they have a tax break until I think 2047 or so.

Utsav Somani: So again, reason why Blackstone is doing this also.

Dhruv Sharma: Who knows? Yeah.

Utsav Somani: All right. That's it from us. We'll see you on Wednesday at four o'clock. Thank you so much.

Dhruv Sharma: Thank you for joining us.