← The Jaeden Schafer Podcast
The Jaeden Schafer Podcast
Ensemble AI Announces $3.3M Seed Round, Co-Founders Alex Reneau & Zach Albertson
01:01:44
Loading summary
Transcript135 lines
- [00:01]
Host
Welcome to the AI Chat podcast. Today on the podcast, we have the pleasure of being joined by the co founders of Ensemble AI, absolute legends. Welcome to the show today.
- [00:11]
Zach
Thanks for having us.
- [00:12]
Alex
For having us.
- [00:13]
Host
Okay, so first off, I'd love if you could introduce yourselves. Maybe you could start and then, yeah, let's get into what you guys are doing and everything that's happening over at the company.
- [00:23]
Alex
I'm Alex, CEO of Ensemble AI.
- [00:26]
Zach
I'm Zach, the coo.
- [00:28]
Host
All right, so tell me, give me, like, just for people listening, give us the brief overview of what Ensemble AI is. What you guys do.
- [00:38]
Alex
Yeah, absolutely. At Ensemble AI, we're working on frontier AI technology and new machine learning algorithms to basically mark the next generation of synthetic data.
- [00:51]
Zach
Yeah, we're coming in and making a suggestion to the field at large that there's more to machine learning that's done today. And then we're actually, with our technology, proposing a new step in the data science in the ML pipeline itself, which has been largely the same for quite a while.
- [01:08]
Host
Okay, so obviously something that's super needed today, but, like, how did you guys stumble upon, like, this need in the market, you know?
- [01:16]
Alex
Yeah, yeah, absolutely. Well, Zach and I go way back. I think we've always kind of connected on, you know, I think the academic side of things and, like, how are we going to bridge this gap between academia and a lot of the great things going on there in industry? You know, we kind of went our separate ways after school and kind of during my PhD experience when I was at Northwestern, I had kind of the opportunity to look at, like, the data quality problem from a ton of different angles. I was fortunate enough to work with a lot of amazing people in that space who are very intelligent and great. And in that we basically stumbled upon this ip and after kind of evaluating how this could be extremely significant, especially in very high risk, complex problem settings, I knew exactly what Zach was doing, and he was the first person I thought to give a call to.
- [02:14]
Zach
Yeah, at the time I was working in consulting, working on some sort of experimental AI stuff. I worked a lot with data scientists, but also I worked a lot with people who knew nothing about data science and trying to deliver client projects, doing a lot of, like, actually tech space stuff. We were working with OpenAI with, like, GPT3 before. It was, like, really, you know, in the news a lot. And I think that when Alex and I would catch up, we both were noticing how much data quality was a problem that was not talked about. Like, there would be, you know, many days in which we struggled, like in the projects that I was doing to actually get ourselves across the line just with collecting as much data as we possibly could. And that was the main barrier and we really didn't have a good solution for it. I think that was your experience in the PhD as well.
- [03:04]
Alex
Yeah, absolutely. I think, you know, the modeling side of things gets a ton of attention. That's very evident right now in industry and also on the research front, I think, in academia. But as you get closer to the science itself, it's very clear that the data is what matters the most. If you have much better quality data, you in reality don't need a very complex model. So when I was teaching students on Northwestern and also doing consulting work in my own research, being able to kind of look at all these different settings, data quality was, yes, 100% the biggest problem, really.
- [03:41]
Host
Okay, so you un. Like you. So you're doing your PhD, you come across this IP, you start working on this, you see this issue and you call Zach. What was like the, what was the call there? It was like, hey, man, like I had this idea, do you want to do this company with me? Like, what was that like?
- [03:58]
Alex
You know, I think a lot of it had to do first with. I always, you know, Zach and I are great friends and we go way back, you know, besides just kind of think he's a great person. That was probably the first reason for. The second reason is he's one of the few people I, I've met who does a very good job at, I think, really bridging the gap as we talked about between, you know, academia and then also industry. And how is this going to apply to real world problems and then be able to communicate that effectively? And that's a very rare thing to find. So, of course, he was the first person that I had to go for this.
- [04:32]
Zach
It was kind of funny in that the conversation that we had about starting the company really started well before Alex ever came in with the iPad, because we actually almost did a different company together while we were in undergrad. We met at Northwestern and we were friends right away, but I started doing data science and it was one of the first times it had been offered at Northwestern and he was doing computer science. And I was like, wow, there's this whole thing, machine learning, I've never heard of this before. And you were like, I am already doing this. So we had a lot to talk about, but I think we both kind of shared this vision that, um, there was a lot of potential that was mostly going untouched. So we actually wanted to do demand forecasting for like food and beverage companies. I had some connections there and we almost did this different company doing that with machine learning, but it was just way too high friction. It was hard to scale. We had a few customer conversations and some offers, but it just didn't seem worth it. We didn't see how it would become anything other than a lot of work for very little like turnout. So we kind of shelved it. But we knew that we worked well together. And then we stayed in touch after that. And he was telling me, I'm working on this research. And he kind of vaguely described what it was. But I remember you were like, I think this could be a hedge fund. If we were to do a hedge fund that you'd be interested in doing that with me. And I was kind of like, maybe. But I think in the back of our minds we sort of assumed the other would be doing something interesting that we could just sort of watch happen. So when he called me up, I kind of went, okay, something must have been happening with this research that you were doing. And I basically just, I think expected you to say, I'm starting a company, I'm going to tell you about it. But that's not what happened.
- [06:13]
Alex
Yeah, I think like, like any big problem, it starts off in a very small setting to begin with. Right. So this started off and, you know, to Zach's point, thinking about it even just like in the trading context, which is a very difficult problem setting, and then thinking about the limitations in those problem settings with machine learning applications and then kind of looking at it at a broader scale like, wow, this data is actually being used in many different industries in context, time series and tabular data in particular. You can imagine all the different use cases such as medicine, outside of finance, lending and risk, many different use cases, you know. So we started kind of thinking about, like, how is this going to scale beyond that? Right. So kind of to why I thought it was a great opportunity to really bring Zach into the picture is, you know, to really bring in the people who one you trust, but also that you connect really well with on how is this going to apply to real world problems outside of this isolated problem setting was a really big thing for me at the time.
- [07:20]
Zach
And I think to that end, you know, I, like I said I was working in kind of applied AI, seeing from a much higher level and like actually doing a lot of coding. I was like working with data science teams. I think that if anything I sort of went, this is too much of a Problem. I don't really want to do this. And initially I told you no, and I was like, but I will help you think through this because I'm in the industry right now. It just seemed like I was at a time where I was kind of like. Like, I've just gotten to this comfortable place in my job. I don't really want to do this thing. I don't want to take on this huge amount of uncertainty and risk. But the more that we talked about it, the more that I started to kind of see through that and go, okay, there is some risk here, but actually, this is really significant. We had a few calls after that, and I think that by the third or fourth one, I went, okay, actually, I will do this, because I don't think anyone else is going to.
- [08:14]
Alex
Yeah. And I think that's kind of where Zach and I kind of set ourselves apart, I think from the masses and the stereo community. And in a lot of ways, it's just, you know, we really believe this is more of a responsibility than just like a nice idea of like, oh, I would love to run a company or do startup. We really believe in working on stuff that we believe deserves to exist. And, you know, obviously being in the PhD, you get to see a million great ideas, probably would turn into good companies, you know, but this is really something that, you know, we both look at and say, wow, this needs to be served properly and taken with huge responsibility. And that's how we're going about it.
- [08:53]
Zach
Yeah. Just to kind of cap it off, I think, like, the reason I think Alex, I trust him to do something like this is because neither of us, and I know Alex well enough to. To count on this a hundred percent, is that we're not in this for an outcome which a lot of founders that we meet unfortunately, are like, I think we're doing this because we think it needs to happen. Basically, what we're trying to do, in short, with that data quality problem, is make all of machine learning, all of AI, significantly better than it is today. And there's a barrier that is preventing it from being what it could be. And we're in a place to solve that. That's pretty much it. Like, the data quality solutions that exist today just are not cutting it. If they were, things would be pretty different.
- [09:36]
Host
What do you think some of the biggest, like, let's say as you scale this, you know, this gets mass adoption. What do you think some of the biggest benefits to data scientists, the globe, the products that we're seeing today will come from this what are some of the biggest, you know, positive outcomes?
- [09:53]
Alex
Yeah, I think besides the, you know, the clear value of using the product itself and creating more machine interpretable data for statistical models to learn from and you know, the value prop that we have on a technical level, I think it's also a thesis and understanding that at where the space is at right now is super nent. And I think like as a company we really try and drive this point home, which is there's going to be new technical innovations coming in. I think a lot of people look at the space right now as crystallized, which we talk about quite often and it's really kind of moving past that hurdle, I think and getting people to realize that there's going to be new things coming in. It's meant to be changed. Right. Because it's so nascent right now.
- [10:39]
Zach
Yeah. A big part of that like this crystallized mindset is it's just a structural element of machine learning and data science is that you have to go through a lot of education in order to do it. For the most part there's a lot of great low code, no code platforms out now and tools like that. But for the most part to really do cutting edge stuff, typically you need to be in a PhD program or you need to be in industry doing like really complex research. And unfortunately that's a huge bias because it makes you look at the field as it is, not as what it's not. You're not looking for this kind of negative space. And so what we hope with ensemble kind of like long down the road is because we address data quality. That means that modeling is much more powerful and you don't have to have all these like tricks up your sleeve, like if you're a PhD to make a model work in a difficult situation. So what that means is we can give someone with a three month data science bootcamp the power to make a model as high performing as a team of PhDs could make and with very little effort.
- [11:41]
Host
Yeah. And I think when you start looking at like rolling, you know, I like the way I see this is Stanford. Last year they made PubMed GPT which was like the GPT of all publicly available medical journals, whatever. And they're like, yeah, this thing's great, but if you wanted this to mean incredible, you would, you know, train one on heart disease, one on breast cancer, like one all these specialized areas. The way I see this is like OpenAI chatgpt that kind of stuff. It's a great general purpose, but we're going to get to thousands and tens of thousands of like, highly specific GPTs built on custom data sets that someone might have exclusive access to data, et cetera, et cetera. And I think a tool like this is phenomenal for that. Where people that previously, you know, this was very inaccessible, all of a sudden becomes accessible to train.
- [12:23]
Alex
100%. And I think to that point, even just thinking about it in terms of generative models, I think generative models, they've been around for a while, but now becoming a lot more popular. Obviously with ChatGPT coming out, the public understands it a little bit more. But that's not even the best model to use in many use cases. I think a lot of people are kind of thinking, oh, like in finance, that's probably not actually a great thing to use, especially when the thing you're trying to predict has an infinite dimensional space. So there will be, I think, a realization coming about here. And to your point, like many models is so much better than one general model, right? I think actually a lot of the general models that are becoming popular now under the hood are many models combined. And that's what's very powerful.
- [13:10]
Host
Yeah, 100%. Even ChatGPT and I think anthropic, right. We're doing this like experts kind of architecture where you have the experts. So we ask a question and then it figures out which of these is actually the best at solving that problem.
- [13:23]
Zach
100%. Yeah. It's sort of funny how like the leaders in the large language model space are moving away from large language models and just moving to language models. That was a distinction we heard from a researcher today. We were talking to him and he kept saying language models. And I think we just instinctively said, oh, LLMs. He goes, no, language models, because LLMs are not good for this use case that we were talking about. And he was one of the first people that I've actually heard like in industry talk about that.
- [13:50]
Host
That's. Yeah, that's fascinating. And I think that's so accurate, Right. It'll be interesting to see, right? Because evidently there's going to be tens of thousands. Like, how broken down can a CHAT GPT get? Like, how many experts can you get in there? That's a fascinating question. Okay, what I would love to know from you. So you have this great idea. How does it work with, just out of curiosity, with the school and your PhD research you're doing, did you have to give any sort of equity to like your research program or anything that was all kept by yourself? I know school, sometimes they do like that.
- [14:21]
Alex
Yeah. So this is totally separate of.
- [14:24]
Host
Okay.
- [14:24]
Alex
Of the university. Yeah. So this is my own private research on my own time with my own resources. Okay, cool.
- [14:30]
Zach
Yeah, that's awesome. We took great pains to ensure that that was abundantly clear before we started the company.
- [14:36]
Host
Yeah, that would not be a fun question to have after the fact. So. Okay. That's fantastic. Phenomenal. You guys get together, you're like, let's make this thing happen. What was your first steps? I understand you raised a pre seed. What was your like, who did you go to first for that? While we're. How was the talking to the process?
- [14:53]
Zach
Yeah, so the first thing that we did actually was to like make the company and get great advice because we'd never done this before. And so we went through a connection to a small like specialized venture law firm that we knew well. And basically we set up a deal where, you know, I was working, you were still in. In school or you were leaving school. And they are actually invested in us in a small percentage so that we could get off the ground with no capital to start with. And then because they do this all the time, they kind of taught us from the ground up how to go about this. Like we had a lot of videos to watch and things to read before we were allowed to like execute our documents, which was great. Which we'd never heard of before. Yeah, like plenty of firms I think understandably are happy to just sign it away and be like figure it out. But the guys that we got, we were very lucky and they lent this huge hand and so then they gave us some suggestions about how to go about this. And so we just went to our network and we just sort of asked around like who is willing to write a big check? Like our check minimum was 100k and which is a lot for like a pre seed but we knew exactly what we were going to do with it. And then also who is going to be someone with a really high risk appetite that's comfortable to lose this money which like really narrows the funnel.
- [16:18]
Host
Yeah, yeah.
- [16:19]
Zach
And then who can we trust to like, you know, not worry basically because this is like such an evolving field. I mean like one week OpenAI would come out with like some new announcement and the field would just explode. And we wanted someone who wasn't going to freak out. And so we really only Talked to like 5 or 6 people to raise that money. And we closed 450k in like a month for precede and we were very lucky. We just got in front of people that I Think this was such a different idea. They'd never heard of something like this before. They're like, I'm willing to take a risk on this and we're very thankful for that. But if we definitely got to put our sales hat on for that moment, because it's like we're two young guys, we haven't actually done anything yet other than be confident.
- [17:13]
Alex
I think also something that was incredibly important to us was bringing people on that knew or understood that they don't know. And I think that's an incredibly important thing in the space with AI right now. And I think in general, like one of the bigger issues that we've seen in venture capital in general, in the I space is there's a sense of everybody wants to act like they know what they're investing in or they understand the space or the landscape, but in reality it's quite different from what they're used to looking at and how it's actually applied and used. You know, obviously we're practitioners, we have a very clear idea of how that's actually done in these settings. So I think to us it was really important to have people who had a good idea of one, the vision and kind of longer term where that was going to fit into the world. And also us that they had conviction and basically us being the team to be able to carry this forward and do right by it.
- [18:13]
Zach
Yeah, I think that it was so important to your point on everybody being clear that they didn't understand because the last thing we wanted was someone trying to tell us how to run the business. And I mean, they're really, they're investing in us at such an early stage. They're not really investing in a business. So first of all, to get that out of the way, make it super clear, I think was good. And then if someone was willing to wait to take a bet like that and basically say, I'll talk to you when you exit, then we knew that they were going to be the right person and that that was very hard to come by. I think we got very lucky for how quickly we pulled that together. It was just very serendipitous.
- [18:52]
Alex
Mm.
- [18:54]
Host
So, okay, so you go, you raise this 450 for the pre seed. What did you put that towards? What was like, what did you do before you actually started going and looking at raising like a seed round?
- [19:06]
Zach
Yeah, well, the. One of the first things that we did was I quit my job so that I could do this a hundred percent. So I was waiting until the money hit the account. Basically to quit. Because it was like, this was all I could think about after a while. Like, pretty soon I was like, I was trying really hard to, like, be a good employee. Like, which I was doing fine. But I think, like, this was starting to consume the back of my mind because we were seeing so much of the news. Like, I was like, are we going to miss our moment here?
- [19:33]
Host
Yeah.
- [19:35]
Zach
So the very first thing that we did was I moved to Chicago. Things lined up in a very funny. Like, where were you before? I was in San Francisco.
- [19:45]
Host
Okay.
- [19:45]
Zach
Yeah, it lined up in a funny way. Like, basically the moment. The money at the account was also the last month that I had on my lease. And I thought I was just going to stay in San Francisco for a while and keep doing what I was doing or maybe even do this remotely. But I think we just talked and being in person, especially for that first year of the company, when you're working so hard, it's just you. We wanted to be in person. And Alex's family is there and his wife is there in Chicago. And so I'm not married, so it was easy for me. And I just was like, I'm not ruining my lease. I sold most of my stuff, and like, two weeks later, I was in Chicago.
- [20:22]
Host
That's amazing. Okay, so you both are, you know, running this company. What is.
- [20:28]
Alex
What.
- [20:28]
Host
What's, you know, the main burn that you, you know, you're spending is this on contractors. I mean, obviously you have to pay yourselves, but like. Yeah, what does that look like?
- [20:37]
Alex
Yeah, obviously paying ourselves contractors.
- [20:43]
Zach
A lot of travel.
- [20:44]
Alex
Yeah, a lot of travel. And like, really, we. I think a big philosophy that we have as business done in person, and I think, like, you know, being able to come in person and interact with people has a big impact on kind of, you know, our ability to communicate the message properly. So a lot of it was towards that.
- [21:03]
Zach
I think that we knew that intuitively, too. And, you know, the cost of travel is not just like, going on trips and, like, we don't stay in fancy places. It's more like someone wants to meet with us. And if we really feel like it's high value enough, like, we're going tomorrow or we're going to night. We've done that before. Um, and we have had an intuition about that being effective. But that was really evidenced when we went to New York Tech Week in October. We had a few early VC calls, and to be honest, they did not go super well. We didn't understand how to message what we had yet. It's Such a technical product. And we were trying to kind of bring it to the right level of abstraction for a VC to understand, but also that they could picture, okay, how is this going to be sold? And we weren't doing super well at that time. It took some time. And we. I got an email from, like, somebody like that bought, you know, bought some data set and had my email on it, and it was like, come near Tech Week. And I just looked at it randomly in my spam, actually, and I saw it was like four or five days away, and I went, hey, I think we should go to this. And for some reason, I just went, okay, let's go. Didn't even look at the email. And I, like, pulled up the. All the events, and I was like, oh, my gosh, there's a lot here that we can go to. And we were not in the mode of really raising money. We were taking these VC calls in an exploratory way. And we showed up in New York, and the first event that we went to, we were like, okay, the air is different here. Not only because it's New York, but we started to realize that we had a lot going for us. Like, we started to realize how different we were from many other startups. And the message that we were just sort of, like, spitballing was landing. So some of our investors that we have now were. Or one of them is the first person that we talked to in New York.
- [22:51]
Alex
Really?
- [22:52]
Zach
Yeah. And, like, we went to this AWS event, and I went and got a cup of water, and I turned around and almost walked into this. This woman who works at one of the funds. And we started talking, and I remember I said something like, we do data enrichment for machine learning in a way that's never been done before. And that got her attention. And she was like, wait, what? And everybody was talking about LLMs at this, like, speech, and it was just such, like, a perpendicular message, and it cut right through. And I think she was interested.
- [23:22]
Host
That's amazing. Okay, so you go to this event, you meet her. I'm assuming you guys kind of start talking about what that would look like for getting some sort of investment. Was that kind of your next focus, or were you still kind of working on your thing and putting this to the side?
- [23:38]
Alex
I think we were still exploring customer conversations kind of alongside talking to VCs. And I think to Zach's point, there a lot of it had to do with still figuring out messaging. That trip for us, New York Tech Week was amazing in terms of just being able to iterate, you know, person after person, how are we messaging this? How are we getting this across to people? And I think the biggest benefit for us is coming out of that trip with a really solid message about what do we do at ensemble, which is very technical but very broadly applicable to everything in AI. And I think also because of the fact that during this time we were seeing, wow, there is really a bubble in the space and everybody's really approaching this from only one angle, making this problem that we're dealing with much bigger than we even thought. So we were just kind of like, oh, my gosh, we need to move quickly on this.
- [24:31]
Zach
Yeah, it was a week of kind of revelations. Not to be dramatic, but I mean, like, we went to as many events as we could physically go to. So we were, like, talking almost 14 hours every day for a week, and it was awesome. But we did not really understand how to raise yet. We didn't really even understand the landscape that well of vc. And we just wanted to see what was out there and get a lay land of, maybe someone will buy this. And it became clear that we were going to need a raise sooner than we expected. Like, we came away from that tripping like, okay, we thought, we have a lot more time. We actually don't. We need to go now. And then it became a little awkward for us to raise because by the time we realized that, it was almost November and, like, that is not a good time to try to raise right before the holiday.
- [25:22]
Host
Okay, so this is what you say, but I do remember talking to you and you told me that Christmas May or like, was it. Was it Thanksgiving was like your. Your secret window where it actually is a good time to reach out, tell the story.
- [25:35]
Alex
It was. Yeah. No, you know, I think. I don't know. I think when you're kind of at it enough, you kind of start getting to this point where you're like, I'm just going to start sending messages to everybody under the sun, and I don't care.
- [25:47]
Zach
Yeah. And I think we'd been. We've been DMing some people and, like, sending cold messages, but not that many. And I think we were just sort of intimidated by the prospect of doing that, because we did. We hadn't done it before. And I remember, like, we've been working so hard for a long time. Come, like, mid December, we basically had been working every weekend, like, since we started the company. And I was like, I have to take a break. I need like four or five days to take a vacation. And so I basically didn't answer any meals, like, or emails. I cleared my calendar and while I was away, I remember you were starting to go a little stir crazy because we've been doing a ton of meetings every day and suddenly the calendar is empty. And I'm looking at my phone and I'm getting like, probably 20 emails a day of, like, Alex setting up these meetings. And I'm like, what are you doing? And he goes, I'm sending like 200 dms a day manually.
- [26:44]
Host
It's all manual.
- [26:45]
Alex
Oh, yeah.
- [26:46]
Host
Oh, my gosh. On, like, just on X or what?
- [26:50]
Zach
On LinkedIn.
- [26:50]
Alex
Everything on LinkedIn. Turns out I've kind of always had trouble with vacation. I'm not very good at it. So instead of taking a break, like, I probably should have. Yeah, I just kind of went a little bit stir crazy and, you know, just was like, you know what? I'm just going to send these messages to all of these people at these different venture capital funds. Probably messaged about like, half of Silicon Valley.
- [27:16]
Zach
Yeah, an amazing amount of people responded because we did this like, the week before Christmas. And, like, once I was back online, I started sending some too. But by then, like, Alex's first wave was like coming home to ruse. I think our conversion rate was like 8% or something for a meeting. It was crazy.
- [27:35]
Host
That's why.
- [27:37]
Zach
And we were getting, like, some of them instant replies to dms and we were like, it's like two days before Christmas. Why are people replying? We're on the phone. And I was like, I think people are hiding from their families. But it was probably the most responsive time we've ever had. Cold messaging people was over the Christmas.
- [27:56]
Host
Break because probably every closed deals before the end of the year or like.
- [28:00]
Zach
In the progress and like, they're all on their phone, so it worked out. I would say, like, if you're going to start cold messaging people, like, a holiday is a surprisingly good time because people say they're offline. They're not.
- [28:13]
Alex
That's so funny.
- [28:14]
Host
So what was your, like, prospect list for that? Were you just, like, going on LinkedIn and you just, like, search top VC firms and just like, hit up everyone on it or, like, what were you.
- [28:24]
Alex
You know, I think, like, you start off more picky than you really should be. And I think that was a revelation for us during the fundraising experience. I think even just looking at investors who say, like, I'm AI focused, I think we stopped focusing on messaging people who are just AI focused because it's a meaningless term. Yeah, it doesn't mean anything at the end of the day, like especially as early as we are, it's about who has conviction in you as founders and kind of the longer term vision of this product, especially the AI space is like so, you know, growing right now and it's not really static.
- [29:03]
Zach
So yeah, when people who say AI like it's kind of tapping error, they mean different things and it's impossible to interpret like the music in their head when they say, oh, I'm an AI investor. They could be investing in LLMs, they could be investing in infra, they could be investing in traditional machine learning, which are all related, but they're really different and what it takes to evaluate, I think whether or not a product even makes sense. The skillset is pretty different. And we've met plenty of data scientists that work at VC funds and some of them are excellent and some of them, we are doubtful that they were really data scientists from interacting with them, which is fine. There's a lot of different ways to do data science as well. You can be a data scientist and never do machine learning.
- [29:46]
Host
Right, Right.
- [29:47]
Zach
Yeah.
- [29:48]
Alex
And it's, I mean, I think the point is too, is like what you see in the news is state of the art research. Not everybody can do that though. And I think like that's a realization when people get really excited about this technology is like everybody thinks that everybody's working on some new algorithm that's coming out, but in reality there's only a few companies that are probably implementing it properly and for real use cases.
- [30:14]
Zach
Yeah. And so I think like on your point of being too picky, I think we just got a much more realistic view of where the market is in terms of sophistication. And that's why we say that the space is so nascent because there is a lot more development just in people's knowledge. There's so much to know. And like we talk about kind of like the standardized data science or ML pipeline, but it's only been standardized up until now because it's been like a pattern that has worked, but it's not truly a standard the way that like there are standards in making steel or something like that. Like it's just not been around long enough. Yeah. And so I feel like that's evidenced in like what you said about really believing in us and understanding the market value of what we're doing, not just the technical value. Because most of our investors are not AI investors. Actually. I don't think really any of them are specialized in AI. Yeah.
- [31:06]
Alex
And I think like the Common theme that we have with our investors is, you know, we aren't building something just for today, we're building something for a longer term product roadmap in the future. Right. This is really a foundation for much more than just focusing on data quality rights, because that's a problem we're all experiencing right now. But in two to three years from now, nobody knows what the biggest problem is going to be from there. So I think we really care about approaching it from being strong on the technical side as a platform for much more.
- [31:39]
Host
Mm, yeah, that makes a lot of sense. So bringing it back to kind of, you know, that whole experience of reaching out to all of these people. So you start reaching out to 200 people a day, you get like a million people inbound. Walk me through, like the process of like trying to do all the meetings, what that looked like, and then I understand things have went fairly well. So talk to me about what you guys are doing now, how. And the progress of the raise today.
- [32:10]
Alex
Yeah, you know, we, I think getting the meetings is one thing. Actually doing them is another. You know, we had definitely our own, you know, raising experience which all founders, you know, have. It's not an easy process, that's for sure. We had our ups and downs, but I think throughout it, it was once again just the more conversations that you have, the better that you get at refining your message. And towards the end of the fundraising process, every conversation we were having was landing very crisp and clear.
- [32:44]
Zach
Yeah. And it separates people who not only understand, like the space very well, but just who's serious. And there were plenty of meetings that we had earlier on, even responses to those DMs where people were taking it to kind of explore. And it's just a huge time waster.
- [33:00]
Host
Yeah.
- [33:01]
Zach
And it's hard to parse until you're in the meeting. Waste in time. And so we just changed tech and we're like, listen, either we're moving forward with you or we'll talk to you later. We said it nicer than that, but that was the message. And I think that we had a pretty crazy, like six or eight weeks of like doing four to five BC meetings every day. And at one point we went to San Francisco and we had like two or three days before we decided to go. And we basically just message everybody that was in the Bay Area, like, we're in San Francisco. Will you do an in person meeting? Way too many people said they were available than we expected. And so we ended up doing like, split up. No, we did it because Part of the way we do it is like we don't have like a track that we play, but we have a good back and forth if we connect with someone. And so we didn't want to like split it up, even though it would have been more efficient. I think we were there four days and we did, I think 18 in person meetings and we did a few calls as well. We were there and it was just so exhausting. But it worked because we were able to see who was really serious and who wasn't. And I think the next week we got our first term sheet, but it wasn't even from someone we met there.
- [34:20]
Host
Really?
- [34:21]
Zach
Yeah.
- [34:22]
Host
Okay, tell me the story. Who was the first term sheet from and how did you encounter them?
- [34:27]
Zach
Yes, our first term sheet was from Amplo and we've been talking to them for a while and it was actually a really good timing because we came off of this trip getting a lot more no's in a concentrated space than we'd ever gotten before. So we were tired from this trip and the flight back. I think we felt really demoralized. We were like, oh, we could be getting into trouble here because the market for raising is deteriorating even as we do meetings. And we got a message from our contact at Ample and he said, hey, are you able to meet like basically right away? And I think we really had our fingers crossed. Like it was kind of a sickening moment, to be honest, going onto the call because we were like, this is either going to be awesome news or this is going to be like really crushing right now. And this guy at Amplo, Sam. Sam is awesome. Huge fans. We've been fans of him from the beginning and he told us the first meeting we ever did with him, he was like, you know, I meet a lot of AI founders and I don't know why, but I think most of them are not going to be successful. I think their products are bs. Like, he kind of like took a huge shit on everybody and then he goes, but for some reason, I feel like you guys are the ones to do it. Which we took as a huge compliment.
- [35:46]
Host
But for sure, I think for both.
- [35:48]
Zach
Of us, we were like, okay, we have a lot to live up to. And he offered us a term sheet and it was, it was the best feeling in the world because I think that was exactly what we needed. And they are participating in a round. They're not our lead, but we were so thrilled to have them because just have someone get us as founders and get what we're trying to do and really Believe, like Sam's line is, I think this is going to be bigger than data bricks. And for us to believe that too is really important.
- [36:16]
Alex
Yeah, so important that, you know, everybody you have under cap table is a hundred percent, you know, behind you, you know, or almost has more conviction that you do. It's because this process is, you know, as a founder, it's super up and down grueling. It's just, it's the nature of, you know, going about it from scratch. And yeah, I think, yeah, with Sam and his team, they were super professional. It was very collaborative. I think, you know, to Zach's point earlier, one of the reasons why we love how having everything be so conversational is just because that's a way that we audit people, you know, who we would be interested in taking money from is like, are you. You know, it helps us basically understand if the conversation is flowing well. We know that you know what you're talking about and, you know, it's a good fit.
- [37:03]
Zach
And I mean, we can put it this way. We have noticeably never had a good conversation in which we open in a very structured environment. And often if that, that means they go, well, I'd like to see a deck before talking to us at all. If they want to see a deck halfway through the conversation, that's awesome. And it usually goes well. But if they just want to see the deck and have us on the call, basically it just torpedoes the whole meeting because we have no rapport with them.
- [37:28]
Host
Right.
- [37:28]
Zach
Because we are much more conversational and I think collaborative, it feels way too transactional. And so it's the same pitch, but the energy and confidence we would deliver it with, it was not a winner for us on either side, for us or them. So we were thrilled. And I think coming back to Sam and his conviction and Apple's conviction, we from there had this total whirlwind where it was like all of a sudden everything reversed and we had other term sheets come in and people were interested overnight. We were like, oh, we might run out of money to. We have more capital than we can take. And then it became a process of, okay, we're going to put our money where our mouth is with people believing in us and really having, I think, that sentiment behind us. And we called everyone that was interested that it said they wanted to put in money and said, if you're not 100%, like on this call in, we're going to give the allocation to someone else. And that cleared it right up.
- [38:27]
Alex
Really?
- [38:28]
Zach
Yeah.
- [38:28]
Host
So did a lot of people drop Did a lot of people solidify?
- [38:32]
Zach
Actually, our problem was fewer people dropped than we expected, really. And so then we had to have some hard conversations, which was basically like, hey, you either need to do a much lower amount or we can't do this. And we had a couple of people do the smallest check they've ever written to be in the round, which was very flattering. And I think it made us trust that they were really in it for us and not to meet some internal mandate.
- [38:56]
Alex
It was really important to us to do right by, you know, these people we've been having conversations with for months. Totally, you know, you know, obviously, like, it is a VC and we're a company, we need capital to grow. But beyond that, it's also looking at the people that we're going to be having, you know, close with us for a long time. So, yeah, it was quite important to us to really do, do right by that process. And, you know, for the people who unfortunately couldn't be a part of this round, you know, we're really grateful to, you know, have great conversations with them. And, you know, there's always the a. Yeah, yeah.
- [39:32]
Zach
Honestly, that was probably other than that kind of feeling of like, we did our best, but we might not make it here. I think we accepted the fact that there was a danger before we got our first term sheet that we may go out without ever bringing a product to market, which that's just the reality of doing a startup. That can happen. Probably the most uncomfortable time was calling people to tell them, hey, we've had this great relationship. We really like you, honestly, but we can't take your money. And some people were pretty mad at us, really. I mean, I understand, but it was the people that were not mad at us that made it way harder because I'd be like, I understand it, like, oh, it hurts. Yeah.
- [40:11]
Host
So, okay, so talk to me about you have Ensemble come in, give you guys a term sheet. You have to do clear up all of this, but they're not your lead investors. So who ended up leading the round? How did you come to that?
- [40:23]
Zach
Yeah, so once we got that from Amplo, we had a few others come in and we got put in touch with Slack Fund or really Salesforce, which Slack Fund is a part of Salesforce Ventures and very small team, but right away we thought they were awesome. And even though it's not like, known for slack to invest in like, AI infra, they don't have a huge reputation from that historically. They were really on it and we were very impressed. Not Only with their team's understanding of the market, but they're also just super professional and they were very efficient. And I think that kind of, to our point about people putting their money where their mouth is and just being straight up, like from first conversation to term sheet I think was 10 or 11 days.
- [41:09]
Alex
Wow.
- [41:09]
Zach
And they said they could move fast and they did. But they didn't compromise any of their integrity along the way of like their process or anything. They just stacked it up. And I think that that struck us. And then like with all the term tricks that we got, we interviewed some founders of the portfolio companies and it just seemed like Slack was exactly what we were looking for. They were going to be there when we need them to be there, but they were also just going to trust us. And it's not that other VCs don't trust you, but there's always a temptation, I think to check in a lot. And I think that we are wired in a way that's hard for us to stay focused on the business if we're having a lot of side conversations, updates to people. Exactly. And it's all well meaning. But I think that in totality Slack just stood out as a right fit for us, for the lead. And I think that like the Salesforce name really carried the day because as soon as we started socializing, hey, Salesforce is our lead, if anything like our allocation problem became way worse, which was awesome. But it really felt like the right decision as soon as we said yes to them. I think sometimes you can sign a deal or like say yes to someone and you don't feel 100%. And that's where we got to, you know, live by our own word, which is we were a hundred percent behind that.
- [42:30]
Host
That's amazing. So you get Slack on, you know, Slack fund, Salesforce on as your lead investor, you have to do some hard conversations and start putting that, you know, back together. What is your, what is your like total amount that you're raising and then what is your kind of vision for that going into the future?
- [42:50]
Zach
Yeah, so we set out to raise 3. We're going to land around 3.4. Sort of finalizing some last minute small checks right now actually. But that's going to give us what we need to go and get really excellent engineering team to take what we have, kind of a functional product and make it really scalable to larger enterprises. A huge part of that honestly is just that we know in software things break and if it's just us working with a client, we can burn Up a lot of time handholding them without a team to support because then that means we can't go on other sales calls. That means we're either 100% with them and or fixing bugs and stuff like that. And so having the resources just to bring on really awesome team to do ML and full stack and build this into a very robust product is. That's where we're going. I think we're already like in conversations with the A already, which is hard to believe because we're closing this round now, but we're going to. We're not thinking about a number there. It's just that that could be a year away already.
- [43:59]
Alex
Yeah, I think. Yeah, exactly as you were saying, Zach. You know, basically our pre seed got us to a point where we have V1 of our product and a ton of developed interest on the enterprise level. But you know, the seed is really going to kind of open the doors to take this extremely scalable product that we have that can be used in pretty much every industry and actually scale it beyond a couple teams. Right. Because our bandwidth is minimal between just us and maybe a set of contractors, we really are kind of looking for that foundational team to, as we've been talking about, really build this foundation for the future and what this frontier technology is going to be and how it's going to fit into the ecosystem at large. And we really, really care about kind of how we interact with our customers, making not just the best experience for them right now, but also how that's going to inform the product that we're developing going forward.
- [44:57]
Host
Yeah, that's amazing. So the talk to me a little bit about, you know, I know you've used contractors in the past. Is that going to continue to be your kind of focus? Are you guys going to be a remote company? Are you going to be an in person company? What's Whereas a headquarters? Like, talk to me about next steps once you guys have this round closed down and really start pushing for this next, you know, phase of the company. What is that going to look like?
- [45:21]
Alex
Yeah, absolutely. I think we both agree on the fact that we would like to find the best talent where it is. So obviously that means having kind of everything be in a remote capacity. We're in Chicago, so it's preferred for us to have people nearby if we can. But we're definitely willing to be flexible if it's for the right people. For contractors. I think our experience with our contractors have been great. I think it's just more so longer term that we're looking for. And I think that just comes differently with full time positions.
- [45:58]
Zach
Yeah. And I think with talent too. It's interesting to be in the Midwest because there's a huge talent concentration in the Bay Area, of course for AI and ML and just software in general. But that comes with actually a lot more cons than I think people discuss. One, it's extremely expensive, so you actually have to go out and raise more, give up more just to have the same amount of staff and you can get really great talent outside of the Bay Area. And there's sort of this weird line in AI where that says that's not possible, which is totally crazy because some of the best talent in ML is in Canada. It's like the opposite of the Bay Area climate wise, but yeah, in France as well. And so it's actually more important to us to try to get some people in the Chicago area or the Midwest because there's a lot of untapped excellent computer science talent for people that just don't want to move to the Bay Area. They don't want to leave their home and it's way less expensive and there's just fewer people doing machine learning in the region. And so that's kind of a lot of edges for us to not only get the best talent but to keep them close by. And there is an ethos to the Midwest that is pretty different from the Silicon Valley ethos about how engineering and software is just done. And it is a bubble in the Bay Area. I mean I still live there, we go there frequently. You can tell when you enter or pass through the layer of thought that happens in the Bay Area around software. It's pretty different.
- [47:25]
Host
Yeah, 100%. And I think we'll definitely continue to see kind of this like dispersing of talent just across as I think a lot of these new startups are kind of starting to take off. I, you know, doing our startup, it's 100% remote. And I know for myself like that is the way that I work the best. I used to, you know, have to do a 45 minute commute both ways. And you know, being able to cut that out not only saves me so much time, but quality of life. And so I think a lot of people, you're going to find some incredible talent that appreciates that. Now I know it's controversial because some people are like has to be in the office and all this but yeah, I tend to, I tend to agree with that same theory. So this is phenomenal. You guys have this incredible product, this Incredible story. You're ready to hit, you know, the, the next level. So I'm super excited to follow along. What I'd love to ask both of you is you've been talking to a lot of investors, you've been working with a lot of founders and other people in this space. What are some predictions that you guys have about AI in the next, let's say one to two years? We've already seen so much, you know, disruption, so much innovation. What's a prediction that each of you could make about AI over the next one to two years? Hot take. Could be controversial. Let's hear.
- [48:35]
Alex
Oh God.
- [48:36]
Zach
Hot take. I think Alex and I don't like when people volunteer their own predictions because some of them we know and we're like pretty sure you don't know what you're talking about. Of course we never say anything, but it's just such a complicated space right now. There's so many variables that even if you do know what you're talking about, it's hard to make a prediction. I think that said said, if there's a prediction we can confidently make, it's that a lot of startups are going to go out of business more than normal, unfortunately. I think there's a lot of hype that is been built up and there is a bubble. There has been a bubble for a while and it will burst, I think. And to be honest, a lot of them just, it doesn't take a genius to look at and go, how is this going to make money? Especially with how expensive compute is. Yeah, like you're depending on something that is not only the compute game is going to change, but also many companies are built on technology that is going out of date by the month, whether it's an LLM or some other new architecture. It's just hard when everything is open sourced. So I just don't see viability for many businesses. And yet they get funded and sometimes they get funded a lot. And Alex and I look at each other and go, guess they know something we don't. But also we're starting to see some of them go out of business. So who knows? I think that if there's a prediction we can be sound and it's that.
- [50:02]
Alex
Yeah, yeah, I definitely agree with the fact that it is very difficult to make any concrete predictions given kind of where the space is at. I think it's super important that everybody remains open minded so that we don't fall into this trap of thinking that it's going to be a certain way and it's not, because that's more likely to happen than not. The one thing that I will say, and I'll speak in terms of where the data space is at a year ago we were talking about this, or a year and a half ago, Even right before ChatGPT ended up coming out, was this idea that models are going to start growing in scale and then there's going to reach a point where they're going to make them bigger and bigger and bigger until they realize, okay, the representational capacity of these models is going to start to plateau. So we're going to start seeing models become smaller but smarter. And that's what we're seeing right now. However, the data space is not catching up. And this is where we really are coming into play. And what we really believe in is the data space is still bigger is better, more is better, but that's not actually true. The reality is we want to come to a place where smarter is better. So I believe that at least in the next couple of years, we're going to see a lot of developments coming into play of less. So how are we optimizing the modeling structure more? So how are we creating the best possible examples that we can pass into these models?
- [51:39]
Zach
Yeah, and I think that that comes back to just like basics of machine learning, which is when you're modeling something, you're representing relationships in data that you already have and you're trying to predict future ones with limited information. Right. And I think we've all heard garbage in, garbage out, like a little too much, actually. But it is true and it's sort of so fundamental it's easy to forget. People have been really distracted by modeling for a while. And it's really interesting because some of the model like developments like the transformer and its broader applications are pretty amazing. Like ChatGPT. Of course, there's lots of others as well that are, you know, more specialized, like we were talking about earlier with smaller models. But I think we also know that hallucinations are a problem. I mean, it's not a great term, but problems on the modeling end, that all the modeling in the world won't really change. All the tuning and hyperparameter changes that you want to do won't fix crap data. And so the problem is that working with data is really, really hard. And so as I think sort of the hype around modeling starts to cool off a little bit and I think there will be some consolidation there, the data side is going to really explode because there is so much that can be done there. It's actually so much that is, I think left ignored that there's a reason that like Mark Andreessen has said that synthetic data is the trillion dollar problem. If you can make synthetic data meaningful and important for the end model, then that's going to be an incredible innovation. And today synthetic data has not really been there. Right. You just learn to make more data with the same statistical properties that you already have had, which is not that useful because you just make more garbage.
- [53:24]
Host
Right. Okay.
- [53:26]
Alex
You're upper bounded with the same quality of the data that you started with. When in reality, like especially in high risk problem settings or any settings that's going to make money or be important on an ethical level, you want better performing models, which you need to increase that upper bound of the data quality to make it worthwhile. Synthetic data is extremely interesting as a concept, especially towards data anonymization. There is a lot of importance there. We just believe that there is naturally new steps involved, which is where we come in at Ensemble to really raise that bar.
- [54:08]
Zach
Yeah. And we talked at the beginning about the crystallization of the view of the space and how there's this negative space of all the things that are not being done right now. And that's where the excitement we think will be. Of course we don't know what it is. You don't know what you don't know. But when it comes to data and specifically innovations around synthesizing new data, whether that's making current methods better, which I doubt, or some new innovation, you know, the unlock there is all the modeling you can't do today with current technology. There's lots of predictions that are basically impossible to make with all the modeling in the world. But that's the negative space. Right. It's all the stuff that's left undone that people go, ah, you can't really easily model that or meaningfully model that in production with great changes to how data is worked within the MO pipeline and innovations there that changes things that used to be unmodelable, become modelable and become profitable to model.
- [55:00]
Host
Yeah. What would you say is one of the biggest misconceptions that people have in AI today? Things that probably most people maybe, maybe like investors, VCs, other people in the market that are. They're getting wrong right now.
- [55:14]
Alex
Oh gosh. I would say the biggest one that I've been seeing, and I think we both agree on this, is using a large language model for everything.
- [55:24]
Host
Yeah.
- [55:25]
Alex
I think people are learning very quickly that is not the right thing to do. I think we've seen this, like as an example in financial services, like, you know, when you're working in problems, you know, financial data as an example. That is a terrible idea. You know, large language models are great with text, text data in particular. A lot of the time, chatbots, they work super well just because of the fact that the text or the context that you're passing into that model is discrete. Right. But when you're moving towards problems that reflect continuous or infinite spectrums, like for example, price in the stock market doesn't work so well.
- [56:07]
Host
Right.
- [56:08]
Zach
Yeah. Another way to say what Alex just said, but in I think more layman's terms would be LLMs cannot do math. They're not set up to do that. Right. They predict basically words or content in sequence. Math is not sequential. I think we met these guys. Do you remember this conversation when we were at the contrary event in New York where they were like students asking us, what do you think would be the next major breakthrough? And I don't remember. We were just so tired and I think that we just like cut right to the chest or to the, to the chase. And we were like, if you can make a machine that can understand logic, like, you'll have probably the most profitable AI in the world. Because LLMs don't understand logic. Logic is not sequential. Right. Like algebra is not sequential. But if you can get something that can understand how to do that in many, many different forms and do it universally, that's pretty amazing. Nothing can do that right now. Right. That's why ChatGPT is great at creating code for some problems, but really abstract problems that are nonlinear, it doesn't do very well with.
- [57:11]
Alex
So maybe like another way of looking at it is generative AI is not the answer to everything. I think as a lot of people are probably posting and seeing, even on social media, it's actually most of the use cases that are being used by businesses are. It's traditional ML. That's what most people are still using. And where there's the most value add, it's still really hard to kind of see where and how generative AI is going to impact certain use cases. And for sure, in certain settings, you may not even need it. And I think there's this itch that people have to kind of make it a part of every process because of the many impressive things that it's kind of brought to the table. But still, you know, you don't need to use machine learning or generative AI for everything. There are certain use cases where you probably don't need it at all. And I think, I think that's an important thing to remember.
- [58:02]
Host
Yeah, I love that. I think that's a really great takeaway. As we wrap up this interview, it's been phenomenal to have both of you guys on here sharing your story and what you guys have done. I would love to get one last golden nugget out of you. What is like one thing that maybe surprised you going through this whole process or one thing that you learned going through this whole process of starting a company, you know, getting funding, growing this thing, building this tech. This is a whole new environment for so many people. This is incredible innovation that's in the field right now.
- [58:32]
Zach
So, yeah, what's, what's a, what's a.
- [58:34]
Host
Lesson that you feel like you learned maybe you can share with other people listening?
- [58:39]
Alex
Yeah, I think for me personally, we were talking about this earlier. It's so important to take moments to pause and to reflect on the journey. I think it's such a fast paced environment. Things are changing and moving very quickly. A lot of highs, a lot of lows. It's so important to remember one kind of where you began, where you started, what the motivations were there, and also just like be grateful, I think, for where you're at and what you're doing and just the amazing opportunity that you have to meet amazing people and to grow in this process and to be able to build something impactful or have that opportunity.
- [59:22]
Zach
Yeah, I think for me, I resonate with that. And an element as well would be like, you have to say the quiet part out loud, kind of. What I mean by that is you do. When you're working so closely with someone like Alex and I, you have to state the obvious again and again and again to make sure that you're always on the same page because you're working so hard and it can be easy to never take a break. And that if you need to say like, hey, I don't think this makes sense, you're going to save yourself so much more pain even in the short run by just being like, ah, I'm sure he gets it. Or like, maybe you need a break and be like, like, what I did before Christmas was like, I cannot do next week. I just can't do this. Because waiting for him to pick up on it would not have gone well. And I think that goes with pitching as well. And like your product, we fell into this trap 100%. And I think a lot of other founders that we meet fall into this trap, which is you want your pitch to seem like the conclusions are obvious, the impact is obvious, and like how it's going to make money is obvious and it's not. And I think that almost no VC that we have pitched to and there are plenty of really sharp VCs that we've met don't get it on first pass. Like, there's just, you got too much in your head that you're trying to get out. And you do have to say, here is why it matters. Here is how it's going to make money. Here's how it's different. Even if it seems obvious and the times that we've not done that, we have a hundred percent not gotten the deal like almost every time. But when we have said it even to people that maybe not we're not a good fit for us, we've gotten way further because we're just on the same page with them and it kind of cuts us some slack. And, you know, we don't need to feel like we're smart all the time just because we're like, oh, we got it. We didn't need to be told. Love that.
- [61:12]
Host
Absolutely amazing advice. Well, it's been phenomenal to have you both on here. If people want to find Ensemble AI, what's your website? I can leave it in the show notes, but how do people find that and how do people find both of you if they want to get in contact or learn more about what you're doing?
- [61:25]
Zach
Yeah, absolutely. So online we're EnsembleCore AI and you can reach out to either of us on LinkedIn.
- [61:32]
Host
Amazing.
- [61:33]
Alex
Please do.
- [61:34]
Host
All right. Well, thank you so much for coming on the show and to the listener, thanks so much for tuning in to the AI Chat podcast. Make sure to rate us wherever you get your podcasts and have a fantastic rest of your day.