 All right, everybody, welcome back to theCUBE's studios in Palo Alto. We're live here at Bill Beyond, made possible by Vast in conjunction with theCUBE. I'm excited. Jeff Denworth is back. And we're going to get the customer perspective from Kevin Weiler, who's the head of infrastructure at Aquatic Capital Management. Kevin, thanks for dialing in. Welcome to theCUBE. Thanks guys, excited to be here. So tell us a little bit about Aquatic Capital Management. What's, would you guys get started? How'd you get started? Why'd you get started? Sure thing, sure thing. So Aquatic is a quantitative trading company. It started as just a QTC. And then we just filled some vowels in to make it Aquatic, but it's the former head quant at Citadel, John Graham. And he started this company with the idea of making a purely quantitative trade and having kind of a more, let's say academic sort of erudite culture. And yeah, we got started in, well, I started in 2019. I think the idea started in 2018, got slowed down ever so slightly by COVID, but kind of picked right back up and here we are. Yeah, I mean, when I go to your website, it's like, it's antithetical to what you would expect from the typical sort of Wall Street firm or the hedge fund grinding, treating people like crap generally, but it's like this culture of almost like research scientists. Jeff, how did you guys meet? Oh, we had a sales team in Chicago that had worked with Kevin in the past. And we heard some rumblings about this team that was moving out of Citadel and we were just fortunate to kind of meet Kevin his nickname is Kiwi, so we call him Kiwi. And yeah, they've just been like, it took a little while to convince them that there was like a new architecture approach that made infrastructure simple. And ever since we proved that out, they've just been a great customer. So it's just awesome to have them on the stage too. And I think Kevin, or maybe I should call you Kiwi too. I think on your LinkedIn, I saw you building the world's best prediction engine, I think you called it. So explain that and what you guys are, kind of what's your mission and I'm dying to hear more about your infrastructure and how you decided to sort of take a bet on a small company at the time like vast. Yeah, sure. So my background was in prop trading which is kind of more emblematic of the trading culture in Chicago worked in that for about seven years and then had kind of a small interlude at another company and was very much sold on the cultural, I would say aspect of this particular company. It was an approach to trading that I would say I was hungry for when I was working in prop trading. The charge I would say in prop trading is just kind of keep the money machine printing as best as you possibly can. And it's just a real grind. Anything that you tried to do that was purely research driven was met with a lot of skepticism. And this company was exactly the opposite. This company is very much driven by kind of pure scientific research. And that's kind of the thesis. The thesis is that we can do a better job than anyone by applying kind of scientific computation methods. When I was working with the sales team from prior companies, we were working with products like Hadoop and we at the time were trying to build something that was very research driven. And like I said, it was met with a lot of skepticism. This is kind of what I've always wanted to do is just a pure research trade and the architecture and kind of data engineering they're in. But you asked a little bit about what we do, how does it work? We take in data, a lot of it from various sources. Market data is of course one of them, but that's not everything. And then we try to apply both, I would say very basic prediction analytics to it and all the way up to very cutting edge stuff. So it's a combination of a lot of different modeling techniques. And after we find something that we like as far as a model is concerned, we deploy it and we trade on it and then feed it back into the system and rinse and repeat. So how's it working? How's the approach working? I mean, what kind of history do you share with us? Yeah, yeah, I can't get too into the weeds on anything very specific, but I am more convinced than ever that what we're doing is sticking and working. We're expanding. So that means they're making money. Yeah, good, making money is a good thing. So you're approaching this as presumably a data problem. Explain your data strategy and I'm interested in your infrastructure. Where does vast fit and what are they enabling you to do that you couldn't do before? Yeah, we were very fortunate to kind of start with vast very early on. We needed a place to put these vasts amounts of data. Yeah, and it was really Jeff that actually sold myself and our team on this and their approach. What we have found with the sorts of data processing that we need to do, and it's a multi-step kind of data processing type problem. You begin with structured, semi-structured and unstructured data all the way across the board and then you kind of have to massage it a little bit and then you try to gain insights from it and then you get to massage it a little bit more and then you gain more insights from it and it's this kind of multi-stage approach and all along the way, you are trying things out and making mistakes and the bottleneck, often we talk in high-scale or high performance computing about whether something is CPU bound or memory bound but most of the time in my experience, you're always IO bound. That's really the big limiting factor at least in this particular space. And so any platform that you can leverage that allows that to go faster is going to pay off and indeed that is exactly what happened for us. We were able to move quickly not because we had the fastest CPUs, not because we had the most amount of memory but because we weren't as IO bound as we would have otherwise been. So Jeff, did you pitch this to aquatic as did you pitch your vision or did you pitch it as hey, we can store a whole lot of data and not be IO bound? The only two people that got our vision before six years into the company's history were our first two investors and subsequent to that, we just stopped talking about it because this idea of a thinking machine was just too far out there for anybody to really grok. And so when I met Kevin and his team, like really what we were talking about is a system that could be standards based which makes it easy to deploy and manage. A system that could be self managing in terms of just being like a software appliance that deploys on commodity hardware that's cheap and is always online. And they were the researchers, right? So they always trying to solve different problems using a collection of tools at their disposal and we were up against open source infrastructure that they were trying to couple together. And open source, the whole world's benefited from it. So the key in our case is not to say, well, that's just not the right approach but just to explain what we can do better. And obviously we were successful with these guys. So can you talk about how you're using AI, Kevin, even generally and I don't know how much you dug in to the vast announcements today but I'm curious as to how you see it affecting your infrastructure and your business going forward. Yeah, so whether you, AI is a very kind of general purpose term to look at a lot of different techniques and you can absolutely apply it to many of the things that we do but the most important thing, and I kind of brought this up a second ago is just iteration time. If you're training a model and you're using AI to do that, the most important thing is to be able to do it over and over and over and over again. Or it's not the most important thing but it's one of the big limiting factors. And so you need to kind of throw a lot of stuff against the wall and see what sticks. And the only way to do that is just to have very, very tight iteration. And so that was the kind of obvious value add as far as vast was concerned right off the bat. I mean, we evaluated several platforms. Like Jeff said, we're evaluating, I have a high degree of confidence in my team and their ability to build just about anything. But as one of the members of my team likes to say these things that you can cobble together from open source technologies require a lot of care and feeding and vast does not require that sort of care. Free like a puppy, as they say. Yeah, yeah, that's exactly right. Yeah, free not as in free beer, but as in the free puppy, yeah. And so, could we have built something that moved as fast as the vast platform did right out of the box? I mean, I think if you gave us enough time, probably, but we didn't have time for that in the beginning. We didn't have DNS and email, like really basic things. And so, it seems like a pretty cut and dry case of like this thing will do what we needed to do without us having to mess with it really at all. But then as time has gone on and, I know Jeff kind of, or maybe it was Andy talked about this a little bit earlier, vast has been very good about taking customer feedback or maybe it was, anyway, whatever. The ability for us to move really, really hard on their system and break it in ways that maybe they didn't entirely expect and be really glad about hearing that is a huge value add for us. You know, there are specific features in the vast platform that we requested. And you know, I don't wanna speak out of turn here, Jeff, but like we're not your biggest customer. I know that. You're our favorite. Yeah, of course. Don't tell the others, don't tell anybody else. Yeah, yeah, yeah, absolutely. So yeah, I think that's been a huge part of the story too is, it's an extremely collaborative engagement. In fact, I think we've also, we've given you some of our internal tooling as well and you've been able to feed that back into your own platform. Sure. So I had a couple of questions. So what did they break? Oh, I don't even remember. That was like three years ago. Do you remember, Kevin? So it is, researchers have a tendency to use computational systems in interesting ways that you didn't expect them to. And one of the things that has come up in the past is just this idea of creating lots of tiny, teeny, tiny little files. And instead of shying away from that problem, which is a classic problem in storage, the small file problem is a classic problem in storage, Vasli into it and said, yeah, whatever, go ahead, create lots of teeny, tiny small files and deliver the ability to do that. But it's not only that, right? As Aquatic's business has evolved, as they're building these prediction systems, the more different multivariant data sources that they can assemble, the more accurate their models can become. So what we see is all sorts of different types of data now flowing into the system. And the key is like, how do you synthesize it together? How do you build the infrastructure that can kind of put this all together in one complete picture? Yeah, it's often the case, I would say that you architect a system for a particular pool of data and the data has certain characteristics. Like I said, it could be lots of small files, it could be lots of large files, that's just one axis that you might be optimizing against. And what we have found is that you just don't really need to worry about that with Vasli. You can just throw a new data set at it and the system will behave in a performant way, exactly the way that you would hope it might. We had cobbled together our own system, I can assure you that would not be the case. And files are one dimension of it, right? And so Aquatic's also early days working with us on the Vas database. And I went and visited Kevin and his team a few months ago, and I remember we were in the middle of a meeting and as we were explaining this new transactional analytical concept, like rolled into one system that's also his distributed file and object storage system, he just started like fist pumping like Arsenio. It's like, that's the first time I've seen an Arsenio fist pump in a long time. Yeah, it was perhaps showing my age a little bit, but also I have a lot of experience with these kind of SQL query engines. And it was very promising to see Jeff and his team speak in an informed way on some of the kind of deeper aspects of those. Like, for example, Prada could push down on Parquet files. This is something that the Vas data platform at least advertises to do kind of out of the box and without you having to think too much about it. I can assure you that architecting that yourself is a very subtle and nuanced problem. And having a platform kind of, again, take care of that for you in the same way that you would hope that it would take care of your, just basic storage needs, it's a huge value add. We've been writing actually last, George Gilbert and I for the last several months now about the future of data platforms. We've been using Uber as an example, who they've in 2015 wrote this extremely complicated system so that we can put together people, places and things in real time. And so to your point, Kevin, not everybody has the resources of an Uber. So you guys are trying to build that horizontal infrastructure. Kevin, what was your aha moment with Vast? Was it sort of when they showed you the data engine and the database, was it before that when they were sort of breaking bio boundaries? I'd say there's a couple of things. I mean, the first thing very early on is, I was, and I think, I don't remember who it was earlier in the presentation talked about this, but I kind of worked as a data engineer during a time when the philosophy was bring your compute to your data. And Vast came along and was sort of like, nah, don't worry about that. And that was a really big deal. It's like, we have these extremely fast networks. You just don't really need to worry about doing that. That was a huge aha moment for me. But then more recently, this idea of not having kind of file storage and database storage be separate and having it kind of scale in a way that kind of makes sense was sort of just the next iteration of that for me. What did that do for your business? I mean, specifically, is just get other things. Yeah, yeah, I mean, I guess that it's allowing us to examine the possibility of having one view of our data or sorry, one place to put our data, but multiple views. So the original Vast platform, you have object storage and NFS style storage. Okay, now we're just adding another one, which is database. Why is that a big deal for us? Researchers like SQL, that's how they think. It's how a lot of people that work in this space think. And you can express things in a sort of declarative way that's more sensible for the problems that you might be trying to solve. Instead of giving a prescription about how to manipulate data, you're telling the system, here's what I want back. This is what I wanna see. That's a very powerful tool for a researcher. And that's what I'm most excited about kind of going forward is putting these tools in the hands of researchers and seeing what interesting things they do that light the entire system on fire. And that's a common theme we see with a lot of organizations, right? If I take object storage or file storage as an example, in the research space, what you have is a bunch of cheap customers, no offense Kiwi, that ultimately would use a database if you had something that was infinitely scalable and supported the needs of incoming streams and also didn't cost you a king's ransom, right? And so what we're now doing is we're realizing that people that have been building their data into these like specific data containers, like everything from parquet to HDF-5, weather codes, oil and gas discovery codes, all sorts of like physics codes, like all these different codes where people create these constructs and dump everything into file systems or object storage. So why wouldn't you just arrange this in an exabyte scale database? Customers like, well, what's the interface? We say it's a SQL, well, what's the downside? Where can I get one? Right, that's the problem, right? That's right. Was your aha moment before or after you started the company? So I, you know, Renin originally pitched me on the idea of Vast when he was trying to get me to come and he said, well, you know, we're building an interesting storage platform and I told him, absolutely not. I'm not interested in this at all. You ran, you didn't walk, you ran. We ended up closing over a handshake where he finalized basically a five-part progression of the product that he was thinking about that ultimately ended up in what, at the time we were talking about is that we actually had the deal that I kind of engaged and said, okay, I'm in. We were just talking about thinking machines for the rest of the night and it just felt so natural. So that's why I'm here and that's the vision that I want to help bring to the world. Very cool. Well, Kevin, thanks for zooming in from Chicago. Thanks Kiwi. Thanks for your time. Thanks, I appreciate it. Thanks guys. All right, and thank you for watching. Next up, PowerPlay, the power panel with the analysts. We've got Rob Streche and Sanjeev Mohan. So keep it right there, John Furrier will be back. This is Dave Vellante, you're watching Build Beyond, live on theCUBE Studios in Palo Alto.