Soumyadeb (00:00:02) - Something you are not excited about and is not your strength is a waste of your resources and time to try to fix that, as opposed to you can just go and find someone who is good at that while you focus on what you can do better.

Ariana (00:00:18) - Welcome to Secret Ops, the podcast uncovering the world of operations one episode at a time. I'm your host, Ariana Cofone, and today's guest is Soumyadeb Mitra, founder and CEO of RudderStack, which is a company that helps data leaders to collect, unify and activate their customer data. Now, Soumyadeb has seen the evolution of the world of technology and data throughout his career and even before that when he was getting his PhD. So we were really able to dive into the evolution of artificial intelligence and machine learning to really understand how it is looking today and why is it so popular today? So let's jump in and learn from him. Thank you so much for joining us. It's so great to have you here. Like I said, I was preparing for this chat, and we don't always get to talk to people that straddle the worlds between technology and data and also being a CEO and a founder of a business.

Ariana (00:01:21) - And they require similar but also different skill sets. So do you mind diving into your journey of how you've gone from being a technical data driven CTO, to now being the CEO and founder of RudderStack?

Soumyadeb (00:01:36) - Yeah. Firstly, thanks Ariana for having me. I'm really excited about this conversation. Now, about that question. I don't think anything was planned per se. And I mean, as an engineer I did my training, did my undergrad in computer science, then came to Chicago to do grad school and then finished my PhD, and then after that went to a startup, spent three years there and then started a company with two amazing co-founders, as the CTO technical guy, and sold the company and wanted to start another company. And it just happened that I I could have gone and found myself like a CEO. And I kind of focused on the technical aspects, but I thought, like, why? Why try to do that? I mean, I can like, uh, be the CEO myself and like, be the founder and start the company.

Soumyadeb (00:02:36) - So it's not like a very thought out thing. But yeah, it just happened.

Ariana (00:02:41) - You jumped in! That's very entrepreneurial of you. Let's just jump in and figure it out. Right.

Soumyadeb (00:02:44) - Exactly. And then and then I'm kind of thankful to like, people that backed you, right? I mean, my investors and so on. So like, they gave me the opportunity. Like often people say they like to go and find a business co-founder but like yeah. So that was not a problem. So here we are.

Ariana (00:03:04) - How is it for you as a, I guess, very technical founder of a business, how do you think your view of building the business might be different from a more traditional, like, you know, business development or more traditional like CEO? Where do you think you've had your unique lens of what you're doing?

Soumyadeb (00:03:24) - Yeah. I mean, it's hard to compare because I've never worked with a CEO who is very different. And even my previous co-founder, he was the CEO, but he was also very technical.

Soumyadeb (00:03:37) - He had an engineering background and so on. So it's kind of hard to compare but in general, I do believe that the different cultures of companies. Right. And you either win by building amazing products or you win by building an amazing go-to-market motion or by being an amazing storyteller. Right? So different companies have different strengths. And often the founders and CEO’s strengths is what you're kind of betting on. Right. And a lot of technology businesses, they have to bet on their product. Right. And we are finally winning based on building the best product. So you have to take care of a lot of other things. But I think in the end you have to build the best product. And that's where I guess, your technical background hopefully helps.

Ariana (00:04:35) - You know, that makes sense. That really makes sense to me.

Ariana (00:04:38) - Because if you're coming in with a certain strength, you've got that lens of how you're viewing that. But then you have to also figure out, okay, how does storytelling, how do all these pieces fit in? Like, for example, you know, I'm an operator, this is what I do. This is what I breathe. I suck at marketing. I'm really bad at it, but I know I have to learn it, to continue learning and doing so. How have you, I guess. What are some hacks that you've learned along the way to round out your skill set as a founder? And especially, again, thinking about storytelling that is not easy. Like you already have the technical piece, which is incredibly difficult to to gain and to get experience in and to be able to do. Storytelling is this kind of amorphous thing that you're trying to pin down? So what are some hacks that you've learned of figuring out throughout the process of creating the company and developing it?

Soumyadeb (00:05:28) - Yeah, I mean, it's a great question.

Soumyadeb (00:05:29) - And it's not that I have figured it out, but like one thing I've learned is that there are things you are good at and then there are things you are really bad at, right? And you can try to improve on things you are bad at. But I've kind of given up. A good example is like, I'm not good at storytelling. That's one thing you probably need to improve because like, finally, like you have to sell like you have to even if you're not selling to customers, you have to sell. You have to sell to prospects. You have to sell to like investors. So there is something you absolutely cannot avoid. But there are other parts like marketing as a good example, right? I mean, there are founders who are super active on social media and then they’re posting in creative ways and they have huge followers and so on. And then I kind of figured out that that's not my strength.

Soumyadeb (00:06:23) - That's not something I get excited about. So yeah, why even bother. But then you need that as a company. But then the good thing is that once you have the resources, you can go and find the right people who can excel on those things. Right. So yeah. So I read this somewhere, but the core idea was that generally the feedback is that you have to go and learn a lot of things, which you probably have to, but something you are not excited about and and you're not is just a waste of your resources and time to, like, try to fix that as opposed to you can just go and find someone who is good at that while you focus on what you can do better.

Ariana (00:07:02) - There's that energizing factor. Like, if it doesn't energize me, yeah, maybe I'll do it for like a couple months and then I'll just sort of drip off because it's just again, like, for me, I don't really like social media as well.

Ariana (00:07:13) - Like I struggle with posting and doing that and it just is not exciting to me. Whereas this stuff really energizes me. Talking to people, learning from people. Right? I could do this forever and ever and ever and that would be different. So it's about that too. Okay, like I gotta learn this thing because I've got to learn it. But also if it doesn't energize me, it's better to find somebody who's an expert in this thing and who really loves doing it to make what we're doing even better. So let's talk about RudderStack, because all we hear about right now is data and how to manage your data, how to leverage your data. And it's really hard to know where to begin. Even as an operator. It's like, all right, where do I begin? In trying to figure out our data pipeline, what we should be caring about, our analytics, our dashboarding. So you all at RudderStack have tried to niche down and to figure out very specific problems to solve for your customers.

Ariana (00:08:01) - So can you bring that to life for our listeners? You know, what is RudderStack doing and why start this company? What problems were you trying to solve?

Soumyadeb (00:08:10) - Yeah. So at a very high level, we solve some of the data challenges around customer data. Right. So any organization has a lot of data, if you take, let's say a consumer company, let's say Amazon, right.They have customers, but then they have people who are sellers and buyers, buyers and customers, but they have the supply chain there’s hundreds of moving parts in that organization. Right. But then the most important. In part, if at all, the customers. Right. And they want to know what their customers are doing on their website and in the app, what are the products they are looking at and, what are their interests so that you can personalize the experience. Right. So that's kind of one part of data that you need to collect and analyze and build applications on top of and to make it better for your customers. Right. So that's what we kind of focus on there like other parts of the data ecosystem, also like the supplier data and other data such as financial data or even third party data that we kind of don't focus on. Yeah. Now why did I start this company? Firstly, I've been in the data space for a pretty long time. My PhD was on data and it was more of a theoretical problem. And on how you secure data and how to learn certain kinds of algorithms on how to secure data, but mostly academic, but at least it was in the broader data space. After my PhD, I went to a company called Data Domain, where again, we looked at data from a different angle though, right? The problem was how to update backups.

Soumyadeb (00:10:01) - So then started another company with two other co-founders, where again, the vision was to automate a lot of things. Marketing has to do with machine learning and so on, right? I mean, if you think about the life of a typical marketing person, there is like left brain and right brain work. Left brain work is around being creative and coming up with amazing campaigns and designing creatives and messaging and so on. But then there is a lot of data work that they have to do, like which customers should be put on what campaigns, what is the next best action to take. And those are the things that you could easily automate with machine learning. And this was like 2014 or 2015 timeframe. Now even the machine learning algorithms like ChatGPT and so on have improved quite a bit. So in ten years a lot of that will be automated anyway. So anyway, that was the hypothesis of my previous company. Like we grew the company, we ended up selling it.

Soumyadeb (00:10:57) - And so again, kind of have been working in this space of customer data and data for quite some time. Felt like an actual opportunity, like in my previous company, after selling the company, we're trying to set up a similar data stack, right? Collecting customer data, building use cases on top of that data, and found that there is an opportunity to do something yet. And here we are.

Ariana (00:11:18) - So here we are. Here we are! Well, it's interesting to me because you've had your eye on data before any of us were even thinking about it. Just like straight up, I think, even going from the theoretical with your PhD into the reality of, you know, growing in selling your startup business before RudderStack. Now, what has it been like to see the evolution of machine learning and artificial intelligence happen to us externally? It seems like it's so fast. To you, I'm sure you're much more knowledgeable of what that's been like, but how has it been going from, hey, we've got this really small circle of experts who are investigating this, researching this into now everybody's talking about it.

Ariana (00:12:01) - Is part of it like, all right y'all finally caught up. Welcome to my life. Or is it exciting that people understand? Or how does it feel to be here at this point, talking about machine learning and AI and how to use data better?

Soumyadeb (00:12:15) - Yeah, that's a great question. Firstly, I won't say like I'm in by no means close to even the first set of people who worked with data. Right. Funny thing is, in my PhD, my advisor and part of my team also looked at high performance and data. Right? So they were running rocket simulations and in these large supercomputing clusters. And this was like pre pre all the big data days. But this was like real, real big data. We are running huge simulations and generating petabytes of data and they're designed this entire enormous system to store the data, process the data. But these are more high like big data at that time.

Soumyadeb (00:12:56) - Right. And there was a whole body of research that was done before, that time frame. Right. One thing that is right is that machine learning has become a lot more mainstream lately, right? Like if you look at, I mean, again, from the early days of computers, like people were talking about, like, yeah, we'll have this amazing AI which will do everything. And then you had pockets of that and you had Deep Blue defeating Kasparov on chess. Right. And those are like 90s. And this was you wouldn't even call it machine learning right now. It is just research and finding the best move because like chess is a very constrained problem. So that was then and then maybe in the late 90s there was this character recognition. Like there was one use case of ML which was put into production for doing like, like postal mails, right? I mean, you could now read it off and print it.

Soumyadeb (00:13:57) - And that was a very, very narrow use case. So pockets of use cases and funny enough my thesis was not on ML but my roommate, he was like working in ML and he said I wanted to get out of ML as soon as possible because there are hardly any job opportunities. Like nobody cares about ML. And then the only people who cared about ML then, were hedge funds. Right. And then he went to a hedge fund. And so it's kind of funny that 2007, 2008, it was the bottom of ML, like nobody really cared about ML. And then like 2013, 2014 was the initial days of deep learning. Right? And this was the paper that Ilya published and there is the standard problem of object recognition, right? I mean, if you have, you're given a bunch of objects and you have to find what objects are there and so on.

Soumyadeb (00:14:45) - And there are standard algorithms for this traditional ML that people use like support vector machines and all the other stuff to do that. And then like the benchmarks were at some level like 50% level, I don't exactly remember, but there is at that level. And then there was this paper from, I think, University of Toronto, where they were using deep learning for the first time and that moved the benchmark by 30 points right from like 50%. They went to like 85%. And this was like a huge jump in ML in the 2013, 2014 time frame. Like this was like the first use case of deep learning and then it has just unlocked like a whole set of applications after that, from that object recognition. Then people went to like speech and like audio, and here we are. It's the same trend that is going on, right? So ML as I said goes through these curves, right? I mean that it is exciting for some time then there's complete disillusionment like people kind of gave up and then now it's like a cycle.

Soumyadeb (00:15:48) - It’s like the hype cycle again and again and again. It's not like one hype cycle. So we are again in that peak of ML, hopefully not a peak. Right? I'm sure. Like we'll keep improving and so on. So yeah.

Ariana (00:15:59) - Well there's this thing. So a couple of years ago, I was a creative technologist, and part of my job was flying around the world and teaching about emerging technology. And everyone at the time was obsessed with blockchain because Bitcoin was skyrocketing and crazy stuff. And part of the educational was that we would go into this really deep dive about blockchain, and then we would talk about machine learning. And people were like, it wasn't even on their radar at all, really. And then it just flipped. Everything was about data maturity, you know, data knowledge. How do you have, like data, all stars in your company, machine learning. And everybody's wanting to know how all of these pieces fit together, which I would say this year in particular, at least in one of the operational contexts.

Ariana (00:16:40) - Now all of a sudden, all of these no-code or low-code tools have all of these AI integrations, machine learning integrations, and we're all playing catch up because like, how do we even begin to start using this thing? How do you recommend people get over the paralysis of like, oh gosh, there's all of these new tools. I don't fully understand the technology, but I know it can help me. And there's that. Like, I don't know that mental block to get over. So since you've been in this world for so long, what tips would you give to people to start learning and integrating it into their work?

Soumyadeb (00:17:15) - Yeah, and I'm not sure. I've been in the space…but specifically ChatGPT is such a huge step forward. Firstly, I was not involved, like there are these LLMs being built in like Google and so on. It is not really accessible. They published a model called Bard, but it was again very like ML researchers mostly using it for very specific things.

Soumyadeb (00:17:38) - But like this, ChatGPT being released to the world is such a huge step forward. I mean, it's like I'm also catching. I'm trying to figure out, like, how can we use LLMs in our like, use cases. So I'm not by any means qualified to give an answer to that. But I think yeah, a lot of the workflows, even a lot of my workflow has changed, right? I mean before, crafting an email was hard. I mean I'm not a native English speaker, so I had to worry about it. Now. It's like, I mean, like ChatGPT usually takes care of that, right? So, I mean, a lot of it is such a huge productivity boost, at least even for me. And even finding information. So yeah, I don't know, like I mean I think but I do believe a lot of things will change over the next five years, but both in the ML community…I'll give you an example like that.

Soumyadeb (00:18:40) - Recommendation systems. Right. People have been building recommendation models for like the last 15, 20, maybe longer, 30 years. And then the whole body of research that has been done in how to build recommendation systems, like Amazon's recommendation, is built on some standard techniques, right, like collaborative filtering and so on. And I believe that entire work might get replaced, like people will start using LLMs to build those recommendation models, like it's, again, deeply technical work, but like there will be change at that level, but there will also be change in how people are writing content and so on. So yeah.

Ariana (00:19:16) - It's true. I mean, I went it was I guess 2016 or 2017, I went to a coding bootcamp and I learned fullstack web development. And now I'm talking to people because they're like, hey, you know, what is your experience after going to bootcamp? And I'm like, your experience is going to be totally different because first of all, all the things that I would spend, you know, hours on StackOverflow searching, you can now ask ChatGPT to spin up a version.

Ariana (00:19:38) - It's so it's so different in how quickly it's changed. I really don't know what it's going to look like in the next couple of years, because part of it has to do with adoption. Part of it also has to do with creativity. Right? Like you with RudderStack, you're saying, hey, let's hone in to customer data. Let's bring insights to our customers to help them see what they're marketing to and what they're using. But that's one very small use case, there's so many things like the healthcare system. I'm really curious to see how machine learning is going to affect the health care system, because I feel like it has the potential to really grow leaps and bounds because it is kind of antiquated. But I think about like even I mean, there's so many different lenses that you can think of insurance like, is there a better way to do insurance and quoting and all of these pieces, is there one area outside of your realm of data and technology that really gets you excited that maybe you're learning too. Is there some piece that you're just like, oh, I cannot wait to see what happens with this.

Soumyadeb (00:20:46) - I mean, healthcare is the big one. I do believe that. My parents are getting old and they have to see doctors and so on, and I kind of make it a point like where if there is some problem, they're going and checking a doctor. I ask ChatGPT what do you think is happening? This is a very small data point. Right. But ChatGPT gives the exact same answer, if not a better answer. Right. So interesting. I think that the knowledge base is already there. I mean for 80% of things you don't have to go and see the doctor anymore. But I think the challenge is still of course, like there are regulatory challenges, ChatGPT cannot give you a prescription and so on. And the second point is about the human. As a human can I trust a chatbot? When you talk to a doctor there is that psychological benefit.

Soumyadeb (00:21:46) - So I don't know how to overcome that. But otherwise like the knowledge is there. I mean, you can get 80% of diagnosis from ChatGPT.

Ariana (00:21:53) - I never even thought of doing that, which seems so silly now because I'm going down Google still, but that makes so much sense. And if you think about it too, sometimes bedside manners aren't really there when you go to see a doctor, and at least with ChatGPT, you're not expecting good bedside manners in telling you something. But it's giving you information. I actually do find sometimes it serves you up information in a nicer way. I don't know, I love my husband, though sometimes he'll serve it up sassy. You know, ChatGPT will give it to me without the side of sass, which always is a good thing, but I also think aggregating, you know, think about how many medical records we have and trying to understand patterns in that data that it's just so hard, even for doctors who are trained, you know, for decades, it's hard to see all those pieces that come together.

Ariana (00:22:45) - Can we get assistance through machine learning, artificial intelligence, to help us get some insights of things that we didn't even think about. Right. I'm curious. I'm very curious about that.

Soumyadeb (00:22:56) - This is the biggest thing I feel like. Aand particularly both in the US and in the developing world, like India, it is a problem in the US it is so expensive to see a doctor in a developing world. It's so hard to find a doctor there, like one doctor per 1000 people. So like in both these cases…I mean, thankfully we can afford. But like, I mean, there are people who cannot afford their doctors and yeah, like they can probably ask like a GPT and get a resolution which they will not have a resolution anyway.

Ariana (00:23:22) - So yeah. Or even give you some inkling of what this thing is or what the symptoms are.

Soumyadeb (00:23:27) - Even which doctor to see. Right. Sometimes you see one doctor and you need to see that other person.

Soumyadeb (00:23:32) - I mean you could have saved one visit at least like ChatGPT would have saved you. So.

Ariana (00:23:37) - Oh for sure. Well this is interesting too. So you spend, would you say, half and half of your time in the States, in the US versus India, maybe?

Soumyadeb (00:23:45) - I’d say like 60/40. Like somewhere around that.

Ariana (00:23:48) - So what use cases are you seeing outside of the US. That would be interesting. Like how are you seeing data being used in different ways within India that maybe we're not seeing in the US.

Soumyadeb (00:24:00) - Yeah, that's a good question. I should probably think about this answer, but I think in some sense India is very much like the US. Right. It's a democracy. Like English is the primary language of communication. And a lot of our curriculum is very similar to how people study in the US, because maybe it's kind of shaped by that. In fact, like a lot of the top engineering colleges were built on the US system of education.

Soumyadeb (00:24:31) - So a lot of things are similar, but at the same time, there are differences, of course, like it's a much poorer country and access to these things. Access to a doctor, which is kind of given in the US, assuming you can pay like in India, getting access to a doctor is hard. Like even if you are well-off, you have to go and wait for hours to see a doctor and so on. So with these kinds of problems, infrastructure has gotten much better. Like from where we grew up. Like language, India has like 30 languages, right. And I'm talking about different languages, like 20, 22 official languages and their like dialects. And so 22 like very, very different languages, it's kind of as different as, like Italian and French and German like that kind of different languages. Although now that everyone…I don't say everyone, but a lot of people speak English as everybody kind of studies English and so on.

Soumyadeb (00:25:24) - So there is that unifying language as well. So yeah, it's kind of interesting. Language is something I think, in ML can play a huge role because like translation and that language barrier will go away with this. Other than that data, I guess one thing India has done well is that we almost jumped some steps. One good example is payment infrastructure, right? I mean, India never built out a credit card infrastructure like the US. But then because we didn't have the money to invest and then and so on. But then everything is digital payment or you could literally pay with a phone and I've never seen that anywhere else in the world. And I traveled so much it's so easy to make a payment in India. And I'm sure that is generating a huge amount of data. And you could probably do interesting things with that data.

Ariana (00:26:23) - I didn't even think about it. So when we're talking about paying with things, are we talking like again for my US brain, we're talking like tap to pay? Are we talking about applications where you can transfer money easily? Like how does that what does it look like?

Soumyadeb (00:26:36) - So you go to a street vendor and we don't have enough street vendors in the US. But like in New York, you have people buying something off the street like somebody's on a pavement selling something. And you could literally scan. They will have a QR code. You scan and you pay, right. It literally takes like five seconds. So that infrastructure, this entire payment infrastructure, and the money is immediately transferred to their bank account from your bank account. So there are no credit card fees or anything. And there's like a zero-cost payment managed by the government and so on. So this payment infrastructure is even better than anything I've seen anywhere.

Ariana (00:27:12) - So would you say that India for the most part is pretty mobile-first? Because we sort of had this computer mid-step situation which is now like when I can't tap to pay things, I forget to bring a wallet. I'm like, I gotta go get money because I, you know, I'm just not prepared anymore. I Covid for sure expedited that transition. At least for myself. I don't think I would have tapped to pay as much but now inserting a card or swiping seems so old school and unnecessary, you know, it's really really interesting. So I can't imagine. When did you start seeing Tap to pay or the QR or the mobile-first? When do you think that's really started to take a hold within India from what you've seen?

Soumyadeb (00:27:59) - Main change was…there is this conglomerate called Reliance, in India, they're one of the giant industrial corporations.

Soumyadeb (00:28:07) - And they launched this thing called Jio, right? It was like a mobile phone service, right. Just like Verizon here. And it was like, I would think, and don't quote me on this, but like I think it is like 1/10 to 1/50 the cost of like data in the US.

Ariana (00:28:28) - It is so expensive here. It's ridiculous.

Soumyadeb (00:28:31) - Exactly. Right. And it was already cheaper in India. And then like Jio launched this 5G service which was much, much cheaper than what was even available then. And that really democratized access that people who are living on $10 a day, they'll still have some kind of a smartphone because smartphones also became cheaper. So that was the big driver like it was in the 2015-2016 timeframe where this cheap data was released. Everyone got that. The other thing was like there was this thing called demonetization. So India, the government overnight made certain currency notes illegal and there was a reason to do that.

Soumyadeb (00:29:19) - A lot of people have accumulated a lot of illegal money in raw currency. And then the government wanted to end that and say, all that currency note is suddenly, overnight illegal. So you cannot cash it out. But it did hurt the economy. But in any case, that was the other reason. Then people felt that carrying money, storing money in cash is not a good idea anyway. So multiple things deep data. And now it is of course convenient. I don't have to carry cash. I just pay with my phone.

Ariana (00:29:55) - Wow. That's interesting because like when I think about, you know, tapping to pay or things like that, there is this painful moment. Usually it comes out of some chaos. Yeah. That all of a sudden you have to learn a new way of being and you're sort of forced out of your comfort zone immediately to have to figure out what to do next.

Ariana (00:30:11) - It's rarely an elegant transaction. Yeah, it never is that way. But the funny part is, I think we do expect that. That we will slowly transition to this thing. And it'll be really…no, like trick technology is hard. It's difficult. I also see that a lot when people are learning new technology, there's so much resistance because it should. Like Apple devices and they make everything very seamless, almost magical. And as beautiful as that is because it's helped with adoption, I've also think it's created unrealistic expectations, like learning. Technology can be hard, and sometimes it will be and it won't be pretty. Even now, I'm a technical person and I cannot get my parents printer to work. Like, I swear to God, I've tried everything and I can not get their printer to work. And that's just part of the muscle. At least when I started learning to be technical, you know, to embrace errors, that was like the first step, right? Like, okay, when you see an error that's like a gift, if you don't see an error, you're screwed.

Ariana (00:31:14) - So having, you know, been technical for so long and having all these skill sets for those who are not technical, is there, I guess something that you can piece of advice that you can give them into ushering them into a more technical headspace?

Soumyadeb (00:31:30) - It doesn't even matter, I think people will be interacting over voice right there, like the whole way you will interact with. Like, look at my doctorate. I mean, when I look at how she uses technology now she has to use a computer because that's cool. Request her to use a computer. But like, her interaction with a lot of things was with Google Voice and so on. And this is like pre LLM, pre ChatGPT. Right? Next year, I mean just interacting with software over voice would be so much more natural. So I don't even know if it will be a problem.

Soumyadeb (00:32:06) - We accumulated some skills around how to set up a computer, which probably doesn't matter anymore. Like you know that the way you–

Ariana (00:32:17) - Like blowing my mind right now because I didn't even think about that. And I should have, because my parents, that was their entry, right? Like, Siri, you know, you know, my dad would be like, hey, the lady said it's going to be 83 degrees. I'm like, what lady? What lady, you know, Siri told me and it's funny because we, you know, I, I learned I remember my parents said, you know, you have to go learn typing. You have to learn how to type. That was like I went to summer school to learn how to type on a computer and how to do it quickly. And that really did serve me. But you're totally right. Now what do you do? You get voice messages from people like on chat. And I'm like, oh my God, there's a voice there. Like it's so disconcerting. But you're right. It’s mobile first, audio first. Right. Like that's what they're just skipping over all the typing stuff that you and I had to navigate for so long.

Soumyadeb (00:33:11) - I do believe that's how Apple gets disrupted, right? I think your phone itself is that they have built that experience, which was amazing. And they're the first to do that, like from BlackBerry to Apple was a huge jump, but the next phone would not even have a touchscreen. It probably just have a I…don't know.

Ariana (00:33:33) - No, that's so true. Like the watch now. Like the gestures. People who have had accessibility issues, right. They've been able to do these sorts of motions.

Ariana (00:33:45) - But for us that don't have that awareness, like my husband all of a sudden is like tapping with two fingers. I'm like, what are you doing? And that's just where we're going, where it's going to be, you know, embedded in how we work and how we live. And we're not going to be stuck to these screens. I'm curious how posture might change with that, too, because we've all got these like, you know, it's just these domino effects. Wow. So seven years old and it's mostly voice and voice text has gotten better too. I think that's also a big change as well. Like meeting notes. I used to take meeting notes and that would be just soul sucking. Because you would miss something. Of course you try your best. But when I try to do voice to text, I would say even four years ago, it would still sucked. Like I had to go through and transcribe it myself and correct transcriptions. And now it's pretty seamless. Like, of course it's still learning, and I'm sure with different dialects and different accents, there's a lot to still learn.

Ariana (00:34:36) - But that's got to be part of the adoption, too. Whoa.

Soumyadeb (00:34:40) - I don't know if you saw that language model, the models that Google launched yesterday in Germany with is multimodal. Look at some of the videos. They are doing voice and image and then text altogether and like you could literally point a camera and then you're drawing something and then you're interacting with that. The experience was around Pictionary and then you're, you're interacting with this computer in all these three models because you're talking because you're giving hints and then you're drawing and then you're also writing, and then it's just amazing.

Ariana (00:35:17) - My God, I have to check that out. I can't believe I missed it. That's incredible. Well, and this also comes to my husband and I like to play video games. And I'm late to the video game party. I didn't grow up doing video games and the first time we ever used VR, I was like, here's the equalizer. So you for over a decade have been on a gaming console. You know, you're so good at being able to use the device. But I, you know, went to dance for ten years when I was younger. I can use my body, you know, I can manipulate and have better awareness of my body than you do with this little device that you're using. And that's, I think, going to be really interesting. The skills that are going to be needed will shift very quickly, and we don't really know what that could be. But for someone you know who's maybe more aware of how they move, that could be more of an advantage for someone who's more into typing. That might not be an advantage. So things that are seen as a negative today, in maybe a year or two from now, it won't matter. Oh my God, you're blowing my mind here. So before we wrap up and learn a little bit more about you as a person, I would love to, I guess pick your brain on one thing.

Ariana (00:36:31) - So for those listening who are operators who are in business. I've got people who are post their careers and just listening to keep Up. What is something that we should keep an eye on? Within the data space in the next, let's say, a year or even six months, what's the one thing that we should keep an eye on?

Soumyadeb (00:36:51) - I think, beyond the ML and AI stuff, like we have been talking about, I think one big game changer that has happened in the data space is this evolution of cloud data warehouses. So ten years ago, it was really costly and painful to collect a lot of data, store a lot of data, and process data like you had to set up a Hadoop cluster if you know what it is. And then there are all these other techs which you have to build. You needed to have a team of data engineers to even set that up.

Soumyadeb (00:37:32) - And today you could literally swipe a credit card and get a Snowflake or a Google, or like an AWS data storage where you can dump data and you can build dashboards or you can even ask text queries like and and so on. Storing processing data has really become democratized, I think. That's one thing, I guess, will change a lot of things as well.

Ariana (00:38:00) - And that's something for operators too because not every operator is coming from a technical background. You don't have to be hyper technical like you said, to be able to use and leverage these tools now. And oftentimes the problems that we're facing as an operator is…we've got so much data across the business. I'm supposed to make a metric dashboard for our OKRs? Okay. How do I hone all this data? Whereas before like you said, get engineering teams involved or learn the skills yourself or, you know, hand roll a solution.

Ariana (00:38:30) - Now you can plug and play. And that allows you to also remain flexible, especially if you're a smaller business or a startup. You know what you're doing today in six months might change. So building out a custom solution is honestly a waste of money. Whereas if you can plug into these the data warehousing that's now easier to plug into. You can experiment, you can iterate quickly. It's like you said, democratizing and empowering, especially for the operators who are listening out there. I found it very empowering as well. And that doesn't even bring up dirty data, oh my gosh, if part of that is cleaning some data and, you know, then I'm in love with whatever tool it is.

Soumyadeb (00:39:08) - Yeah. In fact, there's literally three parts of the pipeline. It was always like that. I mean, you have to collect data and then you have to store it somewhere and process it and so on. And then you have to query that data. And then this entire ecosystem was traditionally like you, you'd buy a tool like Informatica and or run your own ETL pipelines.

Soumyadeb (00:39:24) - Then you use an on-prem data warehouse or a Teradata, or you set up a Hadoop cluster and then connect a BI layer. This entire thing has become so much easier. Now you can connect, use a tool like Datastax or Fivetran or any one of them to like collect that data. You can literally get a data warehouse by swiping your credit card. And then the querying part, people are now investing, building like NLP tools that you can literally ask a question in natural language, English. And then you will get a report back. So like this entire thing has ecome so much easier. Yes, absolutely.

Ariana (00:39:57) - So totally. I remember learning SQL and I'm like, oh.

Ariana (00:40:02) - Right.

Ariana (00:40:02) - Like, you know, if you've never learned how to use or query into a database to get information and you don't know how the data is structured or there's not good documentation, you're just blindly trying to find stuff. It is so difficult that, like you said, natural language processing, all of a sudden the bar just lowers immensely.

Ariana (00:40:23) - And it goes from, I mean, honestly, that stuff used to take me like a day to even get somewhat through it, you know, and to an hour you can start to get some high level insights really quickly. So, man, I could ask you a billion questions, but we've now entered the rapid fire question to learn more about you as a human being. So I'm just gonna shoot them your way and you answer them to the best of your ability. Okay. Ready?

Soumyadeb (00:40:46) - Okay.

Ariana (00:40:47) - Okay. First one is, what is your favorite part of your day?

Soumyadeb (00:40:52) - Morning.

Ariana (00:40:54) - Morning? Are you an early riser?

Soumyadeb (00:40:56) - Yes.

Soumyadeb (00:40:57) - I get up pretty early.

Ariana (00:40:58) - What book are you currently reading or what audiobook are you listening to?

Soumyadeb (00:41:03) - I'm not an avid reader. The one I'm reading right now is Hard Things About Hard Things. It's a book by Ben Horowitz about startup startup culture.

Ariana (00:41:14) - That sounds good. Yeah.

Ariana (00:41:16) - Hard things about hard things. I'm going to check that out. What is the best purchase you've made under $50?

Soumyadeb (00:41:24) - My helmet.

Ariana (00:41:26) - That is super good. That's the first time I've heard that. That's brilliant. Protect that brain. What is your favorite quote? If you have one.

Soumyadeb (00:41:38) - I don't remember who this is from, but it's about startups, right? It goes like this…and I might be slightly wrong, but ballpark. “Bad market, good team, market wins, great market, bad team market wins and great market, great team. Magic happens.”. So market always wins.

Ariana (00:42:05) - Market market wins, market wins. What is something that makes you little-kid happy.

Soumyadeb (00:42:12) - As an engineer? And I haven't done coding in a while, but like when, when you write some code…hard or easy. Doesn't matter. You write something and you, you run it and you see the result that is joy.

Ariana (00:42:28) - Even if it's one plus one and equals two, you're like, yes, two. It happened.

Soumyadeb (00:42:33) - Unfortunately, I mean I haven't coded in some time. And then I was recently writing some code and I was in copilot and so on. So I think the joy of coding is going away. ChatGPT will take that over and many other things like writing. But, but, but.

Ariana (00:42:53) - The boulder feels less heavy, pushing it up the coding hill for sure.

Soumyadeb (00:42:57) - And that means that you feel less joy at the end of it also, right?

Ariana (00:43:01) - It's true. That's actually a really good point. When you press something and a button works, I remember feeling like a god. It's like my button worked and submitted the data. Yes. If you were to go back in your career and talk to your younger self, what advice would you give them?

Soumyadeb (00:43:20) - Don't try to make everyone happy. I think that's maybe the one thing like sometimes like you're often trying to please people and they're not based anyway.

Ariana (00:43:32) - So yeah, that's so true. And a hard lesson to learn. Very hard lesson to learn.

Soumyadeb (00:43:35) - And yeah as a founder. Right. And just add to that, as a founder, I have to make hard decisions and it was hard to let some people go. And it was really hard. I almost felt miserable and I don't know how many people hate me after. I have been a CEO of a company, but probably it doesn't matter, right? So all my life we have tried to we have often tried hard to please people. Probably doesn't matter.

Ariana (00:44:01) - I learned that too. And I say I'm a recovering people pleaser. It's impossible. It's literally impossible. It's a fool's errand. It doesn't mean we don't stop trying. But I feel you on that.

Ariana (00:44:15) - Last question is, what do you want to be when you grow up?

Soumyadeb (00:44:19) - I mean, I'm pretty grown up already, but as I grow older, what do I, I think…Im’ a very inward person. It's kind of contradictory to what I said, but I don't need a lot of, like, external validation, if people think great things about me. Like it doesn't matter. Finally, when I'm dying, I just want to feel that I've done more than I could, what I could have done. Right. So, this was my ability. And then I've done more than that, right? In whatever way. So I think, yeah, I don't know if that answers your question.

Ariana (00:45:02) - Yeah. It does. That's a beautiful thing to think about too. It's not about aspiration, but it's about the feeling that we have at the end of the road here and what do we feel like we've done. This was such a treat. If people are listening and they're like, hey, I want to check out RudderStack. I want to check out everything that you're doing. Where can they find you?

Soumyadeb (00:45:22) - I have a hard first name, so it's pretty easy to find me that way. If you Google me, you'll get to my LinkedIn profile or like it's my first name at rudderstack.com. That's one way. The other is like we have a website rudderstack.com.

Ariana (00:45:35) - And we will link everything in the description if you want to follow along. If you want to test RudderStack, it'll be all in the description for you. This has been so much fun. Thank you for bringing the world of data and machine learning and technology to life. I appreciate your time here.

Soumyadeb (00:45:51) - Thanks for having me. I really enjoyed the conversation.

Ariana (00:45:54) - Great and secret ops listeners as always, thank you so much for listening or watching.

Ariana (00:46:00) - Please remember to subscribe wherever you find your podcast or check us out on YouTube. If you're a visual learner like me! Otherwise we will see you next time.