957: How AI Agents Are Automating Enterprise Data Operations, with Ashwin Rajeeva

25:38So no matter how much data you ask for, right, you can ask us to validate a hundred rows, a million rows or a hundred million rows. We rely on the data plane, which runs in your environment to actually have access to data. And so over the years we've had to build a lot of these pieces. How do you connect them? How do you make sure the jobs are running reliably? How do you authenticate users down into somebody's environment? And so in some sense, we have kind of put all of that together and reused a lot of that stack. But of course, we wouldn't be there without open source technology. So we rely, we built our stuff on Kubernetes, Spark, and all the good work the community does.

26:20and we've also tried our best to you know give back some of it um so we run an open source platform called odp you can actually download it it's like a full data platform which is available for you um you know if if you're interested and that is something which we have built and self-managed because a lot of work our work internal work is also you know making sure we can run different environments at scale that's really cool odp it's just like a github repo it's just a github repo yeah it's a it's our own version of the whole big data management stack and you know it's uh you it's it's open source and we patch it and we build it we make we put the security vulnerabilities there and so a lot of these kind of primitives get used at the underlying system so if you have let's say you know um a rack which is available to you and you don't have compute, you can use this.

27:14This is open source compute, up to date with upstream vulnerabilities taken care of, and it all integrates with the accelerator platform. So it's more of an end-to-end story that we have built. And when you start up, let's say you're four people, you're 10 people, you don't have the resources, capital. So you pick your battles. So we have done the hard work over the years to release different parts of the component. And now we feel with ADM, we kind of have the whole stack. So we have ADM, we have ADOC, we have the data management, we have the compute as well at the bottom. And now we have AI, which can actually accelerate a lot of this work as well.

27:52Jon Krohn:I see. I found the link to your open data platform, ODP. So I'll have that link for our listeners in the show notes and our viewers in the show notes to check out. Thanks for that. So let's go back a little bit to the way that you have built this platform, ADP, to span petabyte scale, complex, hybrid, multi-cloud environments. In a previous interview, you've described this as something like the captain's deck, where you have all the information you need across this extremely complicated system at your fingertips. And you noted that part of the big problem that you're solving with Excel data is what you call data sprawl.

28:34Jon Krohn:Do you want to tell us about what data sprawl is? That's actually not a term that I think I've ever come across. Yeah, it's very simple. I think, you know, I once did a round table and a lot of these, you know, representatives from most enterprise companies come in. And I keep repeating this term enterprise is because I think it's very important to know the kind of customers and the people that we work with. Technology is only as useful if it's applicable. So, you know, it's very hard for us to sell our software to another startup because the startup is, you know, busy building the business. They're not in a space where they are investing hugely in data technology, right?

29:15That's a game which requires money, requires time, and you need to be in a position where you're using the data for something. For example, a customer of ours like a Nestle is a global organization where they invest a huge amount of money in data to provide competitive advantage, right? In different markets and all the skills they sell. Now, it's not true for, let's say, a company like us, which is, you know,$100 million. So some of these problems are, you know, quote unquote, enterprise problems. And one of the things which an enterprise usually does is that they buy a lot of technology. So if you ever have the chance to talk to some of the data leaders, they will always tell you that they have been doing this since the time, 2010, build data platforms.

30:03And so that environment is never homogenous. You won't find a company which is essentially frozen on one technology. And there are different groups in the organization and they use different set of technologies. So now, usually you end up having some cloud environments where there's some data, there's probably some investment on something like a snowflake or data breaks to drive analytics. Maybe some new leader came in and invested in new technology. Of course, there's like a huge legacy landscape of data sitting in data centers. And that's what essentially data sprawl is, right? You have hundreds of terabytes of data on your on-prem clusters, and you moved some of it for analytics into the cloud, and you moved some of it into Google because you signed a cloud transformation journey with Google.

30:57And now you end up in a situation where you have data everywhere. And somebody needs to find out, right, what's where, how do I access it, and maybe put them together in some sort of a report. And that requires you to, number one, know where all the data is, right? And number two is to be able to access it in a good way, which can make sense to a user. And so that's what really data sprawl is, right? It's the nature of, it's the, let's say, eventual fate of any big enterprise that they have data everywhere. And, you know, they want to centralize, but it's hard. It takes many years. Until then, they need to work with all of this data.

31:44Jon Krohn:Fantastic. Well, this has been a great tour of the Excel data platform, all the exciting things that you have going on there, the agentic data management platform, ADM, as well as the ODP. Yeah, open data platform as well. And so I'd like to shift gears here a little bit now to kind of maybe some more generic insights that you might have for our listeners. Of course, it'll still end up centering a lot on Excel data, I'm sure, given how much it's been a big part of your life for the last six years. In an interview, you mentioned realizing that being a CTO is as much of a people problem as it is a pure technology problem.

32:26Jon Krohn:And that you had to move from an architect slash programming mindset to putting yourself in the shoes of customers, of marketers, and employees. So what kinds of guidance do you have for our listeners? What kinds of habits or frameworks should our listeners be thinking of to have the kind of success that you've had as a technology leader and, you know, growing a huge startup like Excel data? Yeah, you know, we've grown from four people to 100 people. I'm just talking about engineering as such, right? We have, of course, more people in the on. And we have done this through COVID and remote work and then coming back to office.

33:09And the reason why I said it at that time, and I really believe it, is because, you know, most of the time as technologists and programmers, you always believe that your version of what you're building, right, is correct, right? It's the best way of doing something, right? So if I build it, of course, they'll come, right? That's the standard thing. And then there are two aspects to it. One is that why do you need to think of it from your customers? perspective because the customers they operate in a different sort of an environment they don't have the same degrees of freedom that you do and most of the time everybody is coming in from some sort of a guidance right so you know the way i think about is that um a new ceo comes in or a new data leader comes in and they come in with you know some sort of vision that hey you know i can do this because i know i've been successful somewhere and now i'm going to go down this technological path.

34:07And then somebody down the organization gets a guidance. And the guidance is, you know, it has two axes. One is the technology, you know, axis of it, which is saying, okay, we need to modernize, we need to use this technology. And for example, now the whole thing is we need to do AI, right? That's the whole thing, something like that. And the other one is the business, which is that there is a business problem you're trying to solve. This is not like a school project where, you know, you build some sort of a website and say, hey, this, you know, Gemini one-shotted this and it's so cool. So usually it's in those axes.

34:40And unless you actually put yourself in the customer's shoes and see if your technology, whatever you're building, is actually applicable to what they're doing, can bring them some value, can make them successful so that they can show this success to their leadership. And then it actually adds value. So next year, when they come back to you, they're willing to buy more or they're willing to continue with your POC, POV, and then convert it into a contract. And then you get revenue and then you expand from there, you build new offerings, right? So unless you think about it, because your engineers are not going to think about it, right?

35:23So you hire people, you give them a task, you say, hey, you're going to work on AI, you're going to work on data, you're going to work on network. Nobody's going to think about it. So somebody in an engineering org has to actually do it. I've always felt that product managers are great, but they're focused on features and issues and bugs. And it's really difficult to be that good at all of these aspects. So as a technologist, somebody in the org, so if not the CTO, then who else, has to think about the fact that, am I building something which will bring value to a customer? And how do you create an org which kind of gets that information?

36:03and so that's from a customer's standpoint and from you know from an engineer's perspective um most people would not be in this is something which i've read as well right at after some point a salary or you know money is not so interesting to people and so if you want to attract um people who want to stay with you especially during a startup's first few years, you don't want people to come and go all the time because it kind of, you know, hinders your ability to carry on context and build at a certain speed. And so if you want to bring in people at an early stage, they have to have the belief of, hey, what is this company going to do?

36:48Why am I doing it? What's my own, what's my way of expressing, let's say, creativity through code or through engineering. And that is hard to do if all they get is like a bunch of Jira tickets which have to be done through Sprint. Right? So you have to create an environment where people think that they are, in some sense, innovating in whatever domain the company has chosen for itself. And for us, that has to do with enterprise and data management. And so the way we want to run our R &D and engineering is to see how much we can innovate here and not just the fact that, hey, the company roadmap has been decided in the beginning of the year and here is a bunch of plans for the rest of the year and do it.

37:35So we have allowed, or let's say, work with our team to innovate on a lot of things. And some of these examples are like the ODP link. This is completely engineer-led initiative. If you don't have a data platform, we needed one internally for testing. We don't want to pay license fees. We don't want to have security vulnerabilities. So we built it. We also built our own vulnerability management systems because we ship a lot of this enterprise software and everybody will scan it. There's a lot of vulnerabilities. So we built a whole AI agent to fix vulnerabilities. And we've built our own kind of data center stack, like an open stack, because, you know, VMware is expensive.

38:14Open stack is too complex. So we built something. And so this is, I think, what inspires people to say, hey, I'm here to do technology. And this company happens to do technology in this arena. And that's where I'll spend time, right? So you got to put yourself in the shoes of the engineer you're hiring and also the customer, you know, you're talking to, and then try to bring in a plan, which kind of works for both of them. So at least that's the way I think about business.

38:41Jon Krohn:That is such a great answer. Sure. You went into a lot of detail there on even the kinds of things I was going to ask about in my next question, which were because you've said in past interviews that a talented programmer is looking for meaning in their day to day that, you know, highly paid engineers still just complain about their jobs. It's funny to me to think that like somebody who gets a hundred million dollars signing bonus at Meta is just like, oh, man. But I'm sure that happens. You know, it's, I don't know what, where the stat is at today, but, um, when I was doing my PhD, something like 15 years ago, I attended a lecture on the economics of happiness, just, just for fun.

39:19Jon Krohn:I just went to this lecture. And at that time it was showing that in the U S if a household was making over something like 80 ,000 or a hundred thousand us dollars a year, that, that is the happiest you can be making more money beyond that point. didn't make people happier. Now with inflation, the numbers are probably a little bit higher. But directionally, I think this kind of gives the idea that you're explaining, which is that beyond having your basic needs taken care of and knowing that you have security for you and your loved ones, the extra money beyond that could end up being a hassle. Yeah, it is.

39:56And it's also interesting, right? I mean, there are studies on developer productivity, right? And you would see that the numbers are insane. I mean, people talk about how an engineer, a software engineer is productive, maybe four hours a day or three hours a day. And the rest of the time is spent in meeting, planning, whatever it is. Now, I feel that even if you take two hours for meeting, you still are leaving a lot of this time out. And if you think about it, what better privilege can someone have than to sit in a usually a great office on a laptop without having to move. Moving is optional, right?

40:35And then get paid top dollar for it. And most engineers then leave their jobs. You know, it's not just, you know, people leave Accelator, but people leave all sorts of companies. And there has to be a reason for it is that most people would be happy because knowledge work is something, you know, which is, which has to do with creativity. It's hard to sit in one place and realize that the work which was presented to you or asked of you could be done maybe in two hours and then you got to sit and find something to do. And it's good for a few days and you spend some time. But after a while, you start getting this feeling that, hey, what am I doing?

41:13I'm supposed to do something better. Let me find a mission which resonates with me and my work and my philosophy of it. And so I think making sure that no matter what the business environment you're in as an executive, the engineers or the R &D teams believe that fundamentally we are in the business of innovation in this field. And there are very few fields, whether it's, you know, something data management, even something as boring as the enterprise content management. I'm sure that there is innovation that could be done, new ways of doing things. And people should believe that they have the freedom to do it.

41:56And they're not just, you know, dictated by quarterly plans and this. I think if we can provide an environment like that, then new ideas come in. And, you know, for us, it's worked out because for a company of our age and size, we have a lot of, let's say, capability that we've built over the years. Whether it's to do with, you know, ODP, ADOC, We have a pulse monitoring system. We have ADM. We are working on the next version of our platform, which will be released in May. And so that is what allows us to do it is where people believe that, hey, in this field of what the company has chosen, data management, there is innovation that can be driven through pure engineering work.

42:40And that's what drives people.

42:42Jon Krohn:Nice. I like that. How do you say in an interview, do you think you have a way of telling whether somebody is going to be passionate about a technical infrastructure heavy mission like data management at Excel data versus somebody who's just coming to collect a paycheck? I think it is, it's easy to tell in some sense. Of course, we've made mistakes there as well, like everybody else. But I feel once you start talking to people, the, and this is what I felt. I mean, I've always felt that management in the technical field can only be done by people who have been in the trenches to some sense. And I'm sure there are models everywhere else and which which are different.

43:35And people have seen managers who work extremely well without actually being on the field. and so the number one thing at least when i look for um potential hires is to see if they can build things right and it's it doesn't have to be working on some data problem right the question is can you build if you are given a problem right and it could be any problem it could be something like, hey, how do you design, let's say, an e-commerce warehouse? How do you design a logistic system or any other business problem? And then can you translate it to something which you have learned? You probably know Go or Python or Java.

44:27Can you put something together which represents a real world problem. I think if you can, then those are the people who bring the most value, who can actually look at like a business problem and then convert it down to what they know. And that's what technology is a good end. Of course, this is for slightly senior people. I think for people who are just coming out of college, it's purely based on potential. Saying, hey, you know, some of it is your scores and your background and some of it, hey, how interested are you into doing this? And then you take a bet and maybe after a few months you decide.

44:57But for most senior people, I would recommend checking if they can build things.

45:04Jon Krohn:That makes a lot of sense. When you're interviewing today, to what extent do you encourage people or discourage people from using LLMs to support themselves? And how do you assess, you know, if you're in a situation where you're trying to assess somebody's capability in Go or Python, yeah, how do you do that and ensure that they're not using an LLM to cheat? Yeah, so one of the things which we've done now, this year onwards, is not to have remote coding interviews, right? We don't want to do it. There are way too many tools which you can use to cheat, and it's very hard to tell. And I always felt personally very uncomfortable, you know, trying to, when I'm talking to you, but I'm always looking for an evidence that you're doing something wrong.

45:52And so it takes away from that conversation, isn't it? And so, you know, we don't want to do that. So the number one thing which we're doing now is face-to-face. The second thing, you know, is that we tell the candidates that, hey, of course, you're going to use LLM to the job. We have given everybody cursor licenses and, you know, got chat GPT licenses, whatever you need. But this interview is all about problem solving. And so there are two things we have done. One is to go towards more practical approach where we kind of created a problem which represents what you would do day to day. so a bunch of failing tests some missing implementation and you won't do it of course you can google right you nobody memorizes all the api so you can google figure out but don't use an llm uh to answer the question and even if you end up using an llm it's unless you install like cloud code or something it's not possible to cheat because you'll have to just copy paste context from different places find an answer um so you know make some sort of a gentleman's agreement with them and then start off.

46:54But, you know, I feel that more and more code will be generated by LLMs. So it might not be a scalable strategy going forward, but this is what we're doing right now.

47:04Jon Krohn:Yeah. Yeah. It is really cool how much you could be getting done with LLMs today. And on that note, a few months ago, you demonstrated an MCP server that prompted a GDPR compliance scoring agent and an interactive dashboard. And it highlighted how AI could complete in minutes what would have historically taken a team days of coordination, research, estimation, and meetings. So as a CTO, building a long-term product roadmap, how do you reconcile this compressed innovation cycle that's now possible today with the slower risk-averse realities of the large enterprises that are your clients? The enterprises, I feel they're all slow in themselves in the decision cycle, but they want it, right?

47:50Because nobody wants to lose out on innovation. So that's why everybody's invested, even though they probably won't be using it, or they don't have the go-ahead from all their compliance and security people or whatever, but they've started using it internally. And so I feel this trend more and more is going to continue. And at some point, you know, some of these people are going to adopt it faster, they will see a competitive advantage, the other people will be like, okay, we are left behind, let's go figure out what we need to do to get there. And over a period of time, it will become more and more easier.

48:29And so our job is also to anticipate the future in some sense, right? We say, you know, we don't build something because somebody wants it today. But you look at where the industry is going and you realize that ai is coming for these sort of jobs let's say as well or these sort of functions more more appropriately and then plan to meet them there right so that's the whole strategy behind it so the whole mcp demo which i did was kind of a idea right which says that um there's this engine which has access to all your data then you have a coding environment now can you connect the two and create something which would have taken somebody a lot of time to do, but an AI can actually do it.

49:10Now, it's not complete, but it's not incomplete as well, right? So the AI has done, say, 60 % of the work, and you could probably prompt it to get 80 % of the work. And then maybe two engineers can finish the rest instead of having a whole team. So more and more, whether it's, you know, vendors and technology builders like us, or it is people who are adopting, you know, AI in enterprises, they will realize that unless they do this, they are going to be left behind. And we want to be in a situation where we meet our customers when they are ready. And that means we have to do all the hard work of making sure we are ready as well with the solution.

49:48Jon Krohn:Great answer, Ashwin. Let's have all of your answers in today's episode. I've really enjoyed listening to you speak. You're a wise man, if you don't mind me saying. It's really nice, really enjoyable and outstanding communication. So yeah, really enjoyed this episode. Ashwin, before I let my guests go, I always ask for a book recommendation. Do you have anything for us? Yes, I do. So, you know, I don't have like a regular reading habit, right? But I try to do at least a book like in a couple of months or so. And I was super interested in understanding about, especially the the g2 as they're calling it right the whole um u.s china rivalry in some sense which is going on so i read a couple of books but i think the most recent one which i'm almost done with is called house of huawei which is by um a lady called eva 2 or eva lu which i forget the name.

50:53And it's all about how Huawei started and, you know, how it became the technology company it is. And also goes in, in the beginning of China's transformation from, you know, kind of a socialist, completely socialist thing to now a capitalist plus socialist society, and also addresses the challenges around some of the human rights stuff. Very interesting, highly recommended for anybody to read, just to get what's going on.

51:22Jon Krohn:Yeah, that's really interesting. I hadn't heard of that book, but it sounds like a good mix of technology, geopolitics, news, just being able to understand what's going on in the world better. Yes, it's a great book. Nice recommendation, Ashwin. Thank you. All right. So for folks who want to get more of your insights after this episode, where should people follow you? Do you have social media accounts that people should follow? I'm a little bit active on LinkedIn about especially Accelerator and what we're doing. I don't unfortunately have an account on X or anything like that. But LinkedIn probably is the best place.

51:54And then Accelerator has some channels. So, you know, please visit our website. And there's our resources and blogs and stuff we have built and written about, which you can read.

52:03Jon Krohn:Yeah, we'll have links to all of that in the show notes. And I got to say, Ashwin, I don't think you need to apologize for not having an X account. People saying today on this podcast that they just have a LinkedIn account is the most common answer. So, yeah, that's how things have evolved in recent years. Yeah, a few years ago it was different, but now it's crazy. It's LinkedIn almost all the time. Yeah, it's LinkedIn also. I feel people are a little measured in communication over LinkedIn than they are on X. And that has to do with how the algorithm and how, what they want the platform to be, which I think is more of like a tabloid than a source of information.

52:47And somehow it doesn't appeal to me. Yeah, yeah, for sure.

52:51Jon Krohn:I think it's interesting, you know, LinkedIn, it's almost because it doesn't really change very much. I think people get comfortable with that. and it's funny like social media platforms they always feel like they need to be changing and adding this and adding that but i think people have found that you know just being able to see you know often business related information that's relevant but often still also entertaining as well and i think a little bit positive so i think something that's interesting like you talk about um the algorithm there over at x it seems like it wouldn't surprise me if on linkedin i am regularly interacting with people who have very different political views from me, but it doesn't matter at all.

53:34Jon Krohn:You don't even know. Yeah, yeah, exactly. And X is different. Yeah. You know, I think it's a personal preference. Some people are that way that they are, they would like to engage in a debate and maybe it's not just all about, and then they find joy doing it. And for more people, for people, it's more of you know, broadcast where they put out their thoughts and then maybe some people find some, you know, something interesting in it. Yeah. But I think one should be able to, and it's something which, you know, I'm thinking about as well, that you should be present as a representative of the company because you don't know where customers are, what people find interesting.

54:15So nothing to do with X itself. I think it's just that some people like me, just not comfortable with that sort of engagement. That's it. But we should be doing more. Yeah, yeah.

54:30Jon Krohn:All right. Well, thanks for that unexpected little social media conversation at the end, Ashwin. I really enjoyed today's episode. And yeah, wishing Excel data all the best. Seems like you guys are on the right track. and yeah, hope to be catching up with Excel Data again soon on the show. Thanks, Sean. Thanks for having me. I had a great time and hope to be back soon. That was an awesome episode. I learned a ton. I hope you did too. In the episode, Ashwin Rajiva covered how Excel Data's agentic data management platform uses AI agents to automate data quality checks, cataloging and pipeline maintenance across enterprise environments, compressing work that traditionally took weeks into hours while keeping humans in the approval loop if desired.

55:15Jon Krohn:He talked about how their Xlake reasoning engine solves the problem of AI models having no context about your internal data by providing tools that connect to data lakes across on-premise cloud and hybrid environments, enabling queries at petabyte scale. He talked about how self-healing pipelines work by having agents detect errors in logs, access the code repository, understand metadata context through the X-Leg engine, rewrite the Spark or SQL code, and redeploy it automatically to compute clusters. That's pretty crazy to me. And he talked about how engineering retention comes down to meaning and innovation rather than compensation alone, with studies showing engineers are productive only a few hours per day when reduced to executing predefined sprint tasks instead of solving creative problems.

55:59Jon Krohn:As always, you can get all the show notes, including the transcript for this episode, the video recording, any materials mentioned on the show, the URLs for Ashwin and Rajiva's social media profiles, as well as my own, at superdatascience.com slash 957. And thanks, of course, to everyone on the Super Data Science podcast team, podcast manager Sonja Breivich, media editor Mario Pombo, partnerships manager Natalie Jaisky, researcher Serge Massis, writer Dr. Zarqar Shea, and our founder Kirill Aromenko. Thanks to everyone on the Super Data Science podcast team for producing another super episode for us today, for enabling that super team to create this free podcast for you.

56:37Jon Krohn:We're so grateful to our sponsors. They allow this show to happen. I guess you guys do too by listening. They're all, we're all working together here. But you can support the show by checking out our sponsors links, which are in the show notes. And if you yourself are interested in sponsoring an episode, you can find out how at johnkrone.com slash podcast. Otherwise, share, review, subscribe, all that good stuff. But most importantly, just keep on tuning in. I'm so grateful to have you listening. And I hope I can continue to make episodes you love for years and years to come. Until next time, keep on rocking it out there, and I'm looking forward to enjoying another round of the Super Data Science Podcast with you very soon.