Hello everyone, Thanks for joining us from where where you are. Welcome to another new episode of Startup Still Say Today. I'm super excited to welcome Vishnu Subra Money into this podcast. Vishno and I have been talking on and off the last three months or so. We first met on LinkedIn through a common post, like many connections these days, and I also had a great opportunity to meet him in person while I was visiting India a couple of months ago.
So Vishnu was super kind enough to invite us to his offices, his studio, his lab which he'll tell us all about very soon and talk about his journey. And I thought it was fascinating in this current world and that we live in where anything and everything we turn on, whether it's news or a podcast or whatnot, everybody talks about AI. And we thought, Okay, why don't we bring Invisioner here and talk here from him about what he is doing in the world of AI and Soshioner. Welcome
to the show. Hi untanny, thanks for inviting me to your podcast. I hope you enjoy the session.
A little intro of me. My father was a banker and I grew up when the banks were getting digitized, so he had a Unix book. That's how my introdiction to computers got started. I kind of went through some of the Linux commands without really practicing. And I've been fighting at my father from probably my third standard to get a computer. And I got one when I finished twelve standard. And I spent most of my colleges in school bunking them and spending in the college library or
in the browsing center on the gaming center. That's where I learned a lot. So I've been in the evolution of a computers much before I understood a lot of things, like I was an early adopter for Yahoo, Male MS and hot Mail. Many people may not understand it at this point of time, but first will understand my first Internet connection was a personal dialup and things are super slow.
From fighting with my father for getting a computer. Now we are running pretty decent sized GP cloud cluster, and we also through our partners, we bring in some of the advanced GPS. So before we get that, let me tell you my journey how it started into AM One day after my marriage, when everyone was watching my marriage video, I was in my Mac and reading through the TensorFlow documentation.
That Google announced Sensorflow one day on my problem on my day of marriage, and I was like thinking that I had a new Mac back then, and having a Mac was a big deal back then in India, not many had it, and I was thinking that, oh, I should be able to run this the TFM and I st on my laptop. I tried doing that and the Mac started crying badly. Was it got very hot. That's how I got introduced to the concept of GPUs and me being a great fan of Terminator movie, I wanted
to get into the AI. Tried understanding couple of things, but dropped it a couple of times, thinking that it's more for PhD, not my kind of stuff, but did not give up. It was still curious and then luckily I came across some of really good content by people like Jeremy Howard from fast A, where some of the things started getting clear. I was able to build things, and luckily on the way, I got introduced to Cagle,
where I started competing in Cagle competitions. Was able to get into the top of the leaderboard top one percent in the world, and luckily got some consulting assignments for companies Fortune and mid sized companies where help companies build their implementation. It was not called a back then. It was usually a lot of deep learning, which kind of morphed into AGE and and all these stuff, but the
basics still remained the same. During this stint, I realized that the GPU's the GPU compute is provided by the hyperscalers are very complicated to use and very expensive. Some of the companies in the US were trying to address this, but they were still expensive for a lot of people out there and countries like India and other places. So I really thought that why not we do it in India and make it cheaper and make it super simple.
That's how the journey started off. My current company called Javaslabs, and the company name was inspired by the Javas from Marvel Movies, which is an AA probably which Sam Adman is trying to build right the AGI out the world, which is like super smart, does everything and all. Is inspired by that. Started this company around twenty nineteen during the COVID. Before the COVID started, the COVID hit us badly. We got stuck in place for more than eight months
without able to do anything. But then things started slowly picking up and now we are doing pretty well.
Awesome, So, mich I mean, based on what you said, I was going to ask you what is your inspiration? You kind of touched upon your inspiration. What I gathered is obviously your dad helped you early on with eunuchs and whatnot, with your first computer. Then looks like your marriage video was the cause of you getting inspired for as well. And then of course terminators. So three people you need to thank for for your journey. I guess right, anyone I missed for your journey?
Yeah, I mean I take inspiration from a lot of people. Here, Jeremy Howard from Fast Andredge carpaty. If it's not for a Stanford open source curriculum, I have not picked up. Yeah right, there are a lot of inspirations the way move on, how to do business, bootstrapped way, how to think long term. I take inspiration from a lot of people.
Yeah. Yeah, that is amazing because I tell myself this as well. I mean, you know you you you cannot shy away from taking inspiration even from a little child. Or there are times when I look at my dog do certain things and I'm like, I mean, if if the dog think that way, there's something in it for us to learn from that as well, you know, so I love.
That absolutely right. The dogs can teach us a lot.
Yeah, So let's let's go back a little bit in your journey vision. I mean, obviously you're having fun doing Jarvis Labs right now, but looking at your profile, looking at what you've done, I mean you've done a lot on the data side of things, right, I mean you know, whether it's HADOO or analytics or even business intelligence for that matter. I mean, so you data was kind of in your ingrained in your blood, or from the beginning
of your career, that's what you were doing. I'm assuming there was some sort of pivot you wanted or something that you didn't like or you loved about data, because I mean, without data, AI is nothing, right. I mean, you know you put in bad data. I mean you have bad data. Your AI is as bad as the data you put in. So can you tell us a little bit about that journey of where you started you know, what what in I mean, where you ended up taking the pivot to this from your data journey, so to speak.
Interesting question. The way I look at it is slightly different. I'll give you some history towards it. My first job was in withprow which is a major I service company in India, and I did not know the difference between services and products, so I was kind of getting bored of each of the projectile was in the pro so I used to fight with my manager and say that no, I'm not interested to get to something else. So I started working with business I think it's sap Bo sap
business objects. Then move to Java swings and applet's building a product for best Buy where probably you guys would sign it that we did some parts of building that, so I have not been to best Buy. Then my last project was MasterCard. I think a lot of things. So what also happened was until two thousand and twelve or something, it was primarily proprilatary software sap IBMB software.
So if you want to get into these, you need to have some account which helps you do it, or you need to get into an expensive training to get it. But the moment open so started kicking in in the form of Android, Spring, Hibernate or how do things started changing. As long as you have a good Internet, that's why we got three G in India, So you do one hundred dollars or Rube reach outs. You get like two GBO three GB connection. Now you can start download things,
a lot of things. Right. That allowed me to do get into the open source and when I was working on business objects, I came across an interesting problem which was also the root for me jumping into her. So we had a business objects report which was running for one of a big client, which started crashing because it was running for more than an hour, and I got curious when Google was able to do a search in like seconds, why does this happen? Why we were not
able to do that. So I namely asked my architect, who was a very smart lady, why is this happening? She was also clueless. But then I started digging deeper into it. That's why I came into this concept called
map reduce. How companies like Google, ads is, distributed computing and all those That's where my interests started and we pro luckily got probably one of the earliest projects in the big data space for MasterCard setting up on one of their largest big data clusters and cloud pioneer in that CLOUDA and Horton works and MasterCard was a super India partner with cloud Era, so we had access to Clouda engineers I learned. I got an opportunity to learn
a lot from them. That's how my journey into data started. And I then started moving out. I moved out of a pro worked in a couple of startups on different things like reporting, and then I worked on an analytics company for some time where we were doing things like Spark and all those things. And Spark was also something super interesting. The problem with Hadoo was it was amazing, but the map produce was too complicated, the Java boiler plate and all this comes. And Spark what it did
is it kind of simplified everything. Whatever you could do in a much bigger final and mapp produce, you were able to do in one single line dot map, dot reduce and do it. Things got simplified over the next coming years. But that's how I jumped into the Spark and then Spark had a subset library called machine learning. That's when I started getting to the came means and random for us, all those things, and we started implementing it for some of our customers in the company, and
that's where the tagle thing also started. By we were able to apply all the open source knowledge that I was gaining all these years on real problems, challenge and learn from some of the best in the world. That's how the pivot happened. And I had to use GPUs in order to compete in Taglar competitions, which needs a lot of GPUs. I had one which my friend luckily got it when he was coming back from you as Titanic. We were not getting it back then. That's how the journey started.
Yeah, yeah, So you spoke a little bit about GPUs and I think it's a nice kind of segue into what Jarvis Labs is doing right now, right, So maybe you can spend two minutes on where you've got the idea or the inspiration for Jarvis Labs, right and what are you trying to offer to home and you know who potentially should be looking at Jarvis Labs, So can you touch on that vision?
Got it? So, as I told you earlier, right when I was working for one of the large companies, they were facing one specific problem. They had like huge service in place with GPS, but the develops team that they had were not able to address the needs of data scientists. So what was typically happening was the data scientists wanted a particular environment. Every time they want installations, they have
to go to develop for permissions. It was the develops was built for a different kind of a beast, not for the air world. They did not know how to handle it, so a lot of compute time was getting wasted. This was in the bigger companies and in the smaller companies it was much different, and for individuals it was much different. It was much harder. Smaller companies cannot even have this develop steam, so spinning up GPU instances, installing
Nvidia COUDA driver was very hard. So I wanted to kind of build a company, build a product which kind of makes this simplified, as I have done that several times for cable competitions on platforms like AABSGCP. Struggle getting GPUs in those platforms, even though I was ready to pay at times, the ex very expensive ones right three hundred, and if you want to get it at that time, you need to wait for pretty long. You need approvals,
which you hardly get. There were one or two smaller players which were like kind of wrappers on these things. They make it easier, but they were very expensive, not suitable for the SMB's smaller companies to mid size companies or for the individuals like us. So that's why the idea came of Java slabs. Can we put twenty eighty TI the gaming cards and then build a software layer on top of it which kind of automates all the
tools that are required for data scientists. So data scientists can just focus on model building or deployment and not worry about the software required for the infrastructure. Pad that Sidea with started the company. After we made some bunch of initial orders, NBDA made it pick to their licensing saying that you're not supposed to use this consumer GPUs in data centers. So we kind of had to take a pivot from that and say that, okay, let's buy
the commercial GPUs. They're a bit expensive, but the story remainder the same. So who can use javaslaps If you're trying to do research, if you're trying to frame your models on MULTIGPU clusters or even on a single cluster with agpus or less GPUs, you don't want to get into monthly, yearly commitments. You want to pages for a minute. You want to deploy your models very quickly and you want to have a friendly tech team to support you in case you'll need something, then probably you should look
at Javislabs. For people that have not really looked into cloud providers, we are an alternative solution to Google Cloud or abus. The way I tell some of the vcs that have met us as GCP is good. They have a purpose. They are like the ones who manufacture BMW casts, but unfortunately the entire world cannot buy a BMW or at Tesla. We need the smaller cast like Marthy Kundai
for different use cases. Also, we offer that so if you're inn SMB spending huge amounts of money, you can just switch to a company like Javaslabs for US and reduce your DevOps cost, increase your efficiency and save a lot of money.
That's a that's a nice USB. So, given given what you're offering, I mean, who is using it today?
Can you?
I mean again you don't have to mention names if you can't, but I mean, at least give us an idea of who is using and what they're potentially using it for.
What they're potentially Okay, let's start with who is using it? Uh So we start Since I was into Cagle. Our initial focus was people who are doing tagle competitions, right, we started that. We had a lot of caglers who actually give us the initial feedback helped us remind product.
But then the problem with Cacalus they are like the competition, so there's a huge demand peak during the end of the competition, and we do not have much crowd using the product in the weekdays or when the competition is not ending. Right, So we kind of started looking at more people colleges, labs or people who are offering online courses, companies who are doing researchers. Some of the top companies both in the US and India use Java slabs for
their research and for their model deployment purpose. There are universities from uers where they want to research on building foundational models or diffusion models or multiple research, right, so they need clusters of eight hundreds or head hundreds. These are some of our potential customers.
That's great, and then how long does it take for them to get started? I mean, is it like an easy plug and play. Let's say they say they want to start using it from today to when they can start using the GPUs and you know, apply a certain model using Jarvis Labs, I mean, what do you thay, what do you think is the run made?
Yeah, so we offer two kinds of uh we or I'll say that we have two kinds of uses. One is, if you want to use the platform normally, you just put your credit card and you can start spinning your first instance and probably less than a minute putting the card thirty seconds and launching instances like fifteen seconds, and you can start building your models, downloading your data sets and all those things. But if I was talking about other use case where university took a large cluster, that's
why you have like contracts in place. Then you have the banking transfer and then we block a huge set of GPUs for them and then they do it. That could probably take days because of the all the formalities that we have to go through. For a normal user, it's usually less than a minute or two. Great.
So the way I understood it, I mean it's almost like versus a custom build sort of thing, right. I mean so that so the first one is more swipe your credit card, create your account, You're ready to go. Your second one is a little more custom use case you need a larger you know, processing power, and larger amounts of data, so they need to work with you to create a custom project, so to speak, and the custom set of GPUs assigned for them. So that's a
that's a nice way to probably differentiate. So you said you started in twenty nineteen, which I mean, where do you see yourself in the journey and you know, has it been as per your vision as to where you are in your journey? I mean, and what are your plans next? With Jarvis Labs, we have.
Come a long way, for sure, and the Aspace sift has changed a lot. So our initial focus was towards a very niche community, very small crowd. The market was not as big assets today, the competition is not as fears as it is today, so a lot of things have changed. People are moving towards contractless GPU compute requirement or even people are moving towards token based billing right, so they don't want to get into some of the
use cases. Do not want to get into I want this XGPO YGP You rather, I want to deploy my model and you just take care of the entire life cycle of that instance. Like what I mean is, let's say you find you in a model which is specific for your use case. You don't want to keep this model running twenty four bus seventy sixty l as you wanted to come up only when there is a request, just like how Aw'slambda works. Right function as a service. That's one of the projects which we are actively investing
in it. We are hoping to go live probably in a month's time, the first version of it, and we're also seeing that huge research is going on what does the next thing right. We have evol platform where we enable companies to do a lot of things, so we are keeping a closed eye on how the industry is evolving and we're also making changes according to that to the platform. Okay, god so I was working My last company was in Bangalore. I used to work in Chennai.
These two are some of the biggest cities in India, which probably most of you audiences have heard of. The challenge with Bangalore is it's a very happening place and it's also very crowded place. You want to move from your house to your office and to and fro, it can take anywhere between two to three hours, even though the distance is very short. That's one reason I did
not want to continue in Bangalore. The second thing is I have been a huge fan of Shi the Remainbow where he has incorporated his one of his offices in a place called ting Ghasi, which is much smaller than Kayamato.
And the idea of that is a lot of product companies in the automobile sector, or the companies which manufacture the rice preparation missions, right, they are in remote parts of Switzerland or remote parts of Europe, different Germany and all right, so you need a calm place for you to focus on building the product rather than getting into the rat race that the city offers you. So that inspired me to come to Kuiamatur. And we don't actually
stay in Kaimatur. For people who are native to Kayamato, they say, oh, you are actually outside or your nearby village. That's how people used to say, oh that village, But it's technically part of Kaimatur. It's a place called Kuepu and it used to be called Miniuti. I'm not sure if it's still called right now. The climate is really good. It almost looks like Minioti. Kim is a very clean,
calm place. People are very respectful here. There's a famous Nisha spiritual place by the tech, yoga, and there are mountains all around. It's pretty close to Western Ghatsan, so it's a very peaceful place away from all the happening things. And it's very closer to banglows like six sevenas to drive, So when I want to meet someone or when I want to go for a meetup, I just drive, have a break, I enjoy the bangalow's nightlife and come back to And staying in Kimator helps us to focus on
the work and forget about everything else. And the cost is also way less than in Bangalore.
Yeah, yeah, I mean, you know, I've been a fan of Committer just personally speaking. You know, back in the day when I used to be in consulting, I mean, my company used to have an offshore facility in Comitor. I used to do a day trip coming into Committer, but leave in the evening to go to Bangalore for another visit. I always wondered, I mean, when would I stay there for a night to spend a little more time. This time, when I came to meet you, I was lucky enough to go around for a little bit cour
to see my friend who got me there. So next time, even more even more time in Coimitto and the suburbs to learn about more. So I thoroughly enjoyed my trip. And I mean I hear all the reasons from you on why you know, especially building something brand new and a product, sort of a mindset and things like that. I mean, you need a lot more peace of mind and quiet that helps you do more with the less with less time than just worry about Oh, let me spend two hours to get to work, let me spend
two hours coming back. And that stress itself is is big. So given that I'm assuming, given you're living in Bangal, I mean in Comito, you get a good night's sleep, but I'm sure Jarvis Labs keeps you up at night as well. On certain things. You know, your roadmap and what to do next. I mean one of the things that you worry about our challenges that you see lying ahead vision.
That's a very interesting question, something very difficult founder to answer. Probably the founder gets worried about it the worst. Then we started Jervis Labs, there was little interest in our product, not many people who are interested in it, so we were like the dominant player and we didn't have much competition. People were happy using us. Then a lot of interest in the VC world came in during the COVID. You would have seen the zero interest rate heavy VC funding wented,
which also went in subsidizing the services we offer. Some of our competitors have subsidized the costs a lot, So that's one challenge that we currently face. The second thing is we're also seeing the industry getting more mature. The efficiency is increasing. We also need the There are too many players today than what were the two years three years back, and more are coming into it. That's typically
one challenge. Of course, we are kind of addressing that by not looking them as competitors rather than we are looking them at as partners. We're partnered with some of them for supplying us with the infrastructure. For example, we don't have a base in Europe, but now through one of the partnerships with one of the largest data centers over there, we are able to offer heat shendreds in
the Europe region. We're also talking to some of the companies in the US, so sooner, in the next one or two months, you'll be able to launch GPUs in a closer to US region also, so we're looking we're trying to take that as an advantage. We're also reduced investing in GPUs because we feel there's abundance of GPS available suddenly. But thanks to all the VC money came in, not just we see real estate money and all this came in. In terms of technology, the technology has been
evolving right now. It's not the small set of people who want to use AI, doctors, artists, lawyers, advocates. In the evening, I was talking to elite lawyers who are trying to bring an AI to ads. Some of their use cases dispute resolutions, automatic can also we are supporting them. So the technology has been changing. There are some misunderstandings of what the technology is capable of, so we are addressing those challenges. So it's an interesting time. A huge
competition has increased. I think over the next few years it will be consolidated again thinking long term, keep lean and survive longer to actually have a better outcome.
You mentioned we see is a few times. I mean in terms of where you are in the stage of the company, right, I mean, can you fare a little bit about you know, I mean, are you you know, just in seed stage, waiting a little bit before you go for Series A or Series B. Or what are your plans in terms of, you know, even getting investors into the company. I mean you might say you don't want investors, but I mean just for anyone listening and interested in Jarvis.
Labs, I mean, what is your message to them? So we are a bootstrap company. We have not is an external funding. In the initial few years, we had opportunities to talk to vcs, but they did not turn postive. But right now we make enough, we don't need to raise money and we want to remind that way. I also believe that remaining bootstep offers a lot of freedom and I don't want to run behind millions or billions
of dollars. As long as we make enough and as long as we grow organically, we are happy with it. One of the advantages of staying in a place which is far away from the Silicon Valley of India is also that you don't need to be part of the address. Right we love building products. We will continue doing it, but we don't want to run behind the money. We will make money, but we don't want to run behind it.
So we don't want to go behind we see money because that it's like taking artificial way of scaling up. We want to see how we can take inspiration from companies like Zoo zero from on how we can grow organically, maintain a healthy lifestyle and build a long lasting company then a company that is built to be sold to someone else. So we are not raising funds at least as much as I can see in the future.
That is super inspiring vision. I mean I wish every founder can say that, right, But I mean obviously you see you see you said rat race as well a few times. It's always a question of comparing one to the other. You know, what's my evaluation? You know all that sort of thing. But you you offer a very refreshing perspective, and I love the references you always make to Soho and others.
I mean, I know a bunch of guys from.
As well and understand their culture, how they operate. I mean, if anyone listening here has has not heard about Soho, I mean that's a super interesting case study and a big inspiration someone who comes from our state, Vish News and mine which is Tamil Nadu. So go look up Zoho as well. So probably the last question for you is you you worked in a big company, Like you said,
you know you worked in smaller companies as well. Now you're a founder, you have a team, you have mouths to feed and build, you know, paychecks to give and all that. So and in today's generation, you know, probably different from twenty years ago, where people wanted a paycheck to take home, they wanted stability. I mean, there's a lot more kids and college students trying to be founders, you know, trying to be start their own company, bring
their dream idea to life, and things like that. So having been through this, I mean, what what what are your learnings or your advice you would give to somebody who's in the third year of college or a fourth year of college trying to come out and be a founder. What are your nuggets of advice or three nuggets of advice?
I would say, don't wait till you finish your college, right because when you're in college, you're lucky enough for your father to be for your mother to be funding you your life. So you don't need to maintain your lifestyle, and you don't need to go and convince someone to work for you or work with you. I prefer work with you. You can convince your friends to work with you. So pick up an idea and a lot of things have changed in the technology space. There is a concept
called ind hackers. While you can build product and with tools like GPT cloud building products have become a lot easier if you put enough amount of time and if you kind of read so you can actually build product before you finish your finalier. And if you're lucky or if you're planning enough, you'll be making money also in the way. So by the time you come out, either
you have the experience building the product or lucky. The best case scenario is you have built a product and you also made money, right, so then you'll have a better way to see whether you want to continue with that journey or you will know if it's not your cup of team. Being a founder is very hard. Before I started Javas Labs, I have heard a lot of talks oft from cred who also is a founder of I think free Charge, he sold that or Snapdean or
something like that. He says founder's journey is a lonely journey. I used to think, oh, that's maybe it's not me, maybe it's a still But four years into it, I can say that's true, right, because you have a lot of things going on in your head which she may not be able to share with everyone, or you may not even have the time to share with people, right, So it's good. It's good to try it in your college days and see if you want to do it.
For people who come out of college, I would say, try to get into a particular space which is usually upcoming, pick up a domain knowledge, and see how you can solve a problem differently. Watch a lot of videos of y combinators. There are a lot of videos of See the Way boves Zero. The learn from these people so that you will form a kind of insights on what you want to build, and probably that's a good way to start. Don't be dependent on VC money to start.
That may or may not happen even if you want to. Just a small percentage of companies raised V see funding. A lot of college people I have spoken to, they expect that I will get funds to build this product through VC. Usually we see companies don't fund for your ideas unless until you have like successful exists or you have a very different background.
Those are great pieces of advisation. I mean, I wish I'd spoken to someone like you when I was in college. You know, my first first task when I was in college was to find a job, settle down or go to my ms. But times of job definitely change, right, So your journey is super inspiring, not just for the next generation but even for people like me. I thank who've been working for plenty of years. And if you have an idea, if you have an inspiration, just go
do it. Is you know, is what I learned from you. So maybe one last question and we can wrap up this question. I mean, because you're an AI there's a lot of talk these days about whether AI is just another this is just another bubble waiting to burst, or this is for real this time because of all the you know, money that is being pumped in. There's other stories about oh there's AI is way is waiting for a problem to solve?
Right?
I mean, is technology looking for a problem? Having been in this space much longer than you know, at least me and many others, and you know you've put in your days and nights into AI building this company, I mean, what is your general take on the state of the industry right now and where do you think it's going to go?
So AIR is definitely not a bubble, right whether all the companies all the VC funding will make a profit. Maybe not. But let's say for hypothetical scenario where all the funding for AI and all the research work for a stops right this minute. Right even then, instead of have needing one hundred people team, you can build a lot of products with a small three or ten people company can still do the same. It revolutionizes education. For example, in Indian space right you have like for thirty or
sixty students one teacher. So with AI that can completely change if we don't need much advancements in here. Of course there's a cost involved in it, but assuming that the cost, engineering is something easier, easier to solve than going towards the AGI. So I don't think it's a bubble. A lot can be done with We have just scratched the surface. That's how I see it. It's not searching for the problem. It's we have to figure out how to solve the problems that we have, rather than solving
the fanciest problem. If you're trying to see innovative ways of how to use it, probably skies. Will it be agier, that's a completely different thing, which I don't agree and I don't want that to be true.
Okay, awesome. That's a great insight and a nice perspective from someone who is breathing this day in and day out. So thanks so much for that ution. So let's wrap it up here. I sincerely appreciate you taking the time in the evening time to jump on this and share your insights about Jarvis Labs. I'm sure we'll do another episode, say six months down the line, see looking at your growth and what else you've learned, but wish thanks so
much for joining us. All the best to you and Jarvis Labs and your team and we'll stay in touch.
Thank you, Anthony, all the best for you too. Two