Slater Victoroff founded Indico Data in his dorm room at Olin in 2012. Two years later Indico had gone through Techstars and was VC-backed. Hear the story of how making deep learning accessible to enterprise customers led to massive growth.
Highlights:
Sal Daher Introduces Slater Victoroff, Founder and CTO of Indico Data
Indico Was Founded in a Dorm Room in 2012, Two Years Later They Raised $3 million in an Oversubscribed Round
Machine Learning Is Programming with Data Rather than Code
Deep Learning Is Characterized by a Focus on Unstructured Data i.e., Text, Audio and Images
Indico’s Original Aim Was to Make Deep Learning Accessible to Software Developers
Slater Victoroff Founded Indico Because Most Innovation in Deep learning Was Driven by Industry
Indico’s Experience at Techstars
Semyon Dukach Saw Promise in Indico Others Did Not
“...Techstars really helped us understand was that we were real.”
Why Slater Victoroff Gave Up the CEO Role
The Shift to Enterprise Clients and Massive Growth
Stellar Customer Satisfaction Metrics Was the Secret
Slater Victoroff Was an Unlikely Founder
About Hiring Machine Learning Talent
Slater’s Work at .406 Ventures
What Sets Olin College Apart
How Textbooks Empower the Self-Directed Learners
Why Slater Victoroff Does Not Invest in the Life Sciences
Parting Thoughts from Slater Victoroff
ANGEL INVEST BOSTON IS SPONSORED BY:
Transcript of “Indico Data”
Guest: Slater Victoroff
Sal Daher: I'm really proud to say that the Angel Invest Boston Podcast is sponsored by Purdue University entrepreneurship and Peter Fasse, Patent Attorney at Fish & Richardson. Purdue is exceptional in its support of its faculty, faculty of its top-five engineering school, in helping them get their technology from the lab out to the market, out to industry, out to the clinic.
Peter Fasse is also a great support to entrepreneurs. He is a patent attorney, specializing in microfluidics and has been tremendously helpful to some of the startups in which I'm involved, including a startup came out of Purdue, Savran Technologies. I'm proud to have these two sponsors for my podcast.
Sal Daher Introduces Slater Victoroff, Founder and CTO of Indico Data
Welcome to Angel Invest Boston, conversations with Boston's most interesting founders and angels. Today, we are really privileged to have with us Slater Victoroff. Welcome, Slater.
Slater Victoroff: Thanks so much for having me. I'm happy to be here.
Sal Daher: Slater is the founder of Indico Data, an amazing startup, which I had a chance to invest and I just didn't try hard enough. They were 3X oversubscribed when I saw them at Techstars. I was busy with some crazy stuff with my buildings and I never managed to write him a check. My friend Michael Mark wrote him a check. Super very happy with what they did. Anyway, so at least, we get to talk to Slater and get a little bit of his wisdom, of his founding story. Slater, you founded Indico in the dorm room of your Olin College dorm.
Indico Was Founded in a Dorm Room in 2012, Two Years Later They Raised $3 million in an Oversubscribed Round
We'll talk about Olin later on. That's an amazing story. It's a very Boston thing. Tell us what problem were you solving at that time that you needed to start in your dorm room.
Slater Victoroff: I think the story of Indico and the problem we were first solving really starts with a quote that I sent to one of my professors. This was in 2012. I will never forget this as long as I live. I said to one of my professors, "The war is over, deep learning lost." Now, I was a sophomore at that point. I had just gotten my name listed on a patent and I had had some good initial forays into AI research. Of course, at the time, I knew absolutely everything. Now, I think I would characterize that sentence as the most wrong I've ever been. Maybe I'm just at least happy that I got it out of the way while I was young.
Sal Daher: Slater, I'm sorry to interrupt, but can we just translate for the laypeople, the people who are in the life sciences, deep learning. Explain just what deep learning is, what sets it apart from the other types of artificial intelligence.
Slater Victoroff: That's exactly where I'm going next. I think I'll try to give it without giving a ridiculous amount of context. Deep learning is not a panacea. It's not something actually that's applicable to all sorts of machine learning. Maybe to contextualize some of these classic terms, AI is the super term. AI is anything where you're talking about something that's done by humans today, that computers or machines of some sort will do tomorrow.
Machine Learning Is Programming with Data Rather than Code
Within that, machine learning is a very powerful technique, specifically for saying, really that supports programming with data. That's how I would depict it. Rather than saying explicitly in code, "This is when you turn left. This is when you turn right," you have the ability by using machine learning techniques to say, "I'm going to define when you should turn left or when you should turn right based on this data," and then saying, "You should choose the optimal strategy."
Sal Daher: Instead of a programmer writing a computer program to tell the computer what to do, you give the machine learning algorithms the data and it sorts through, and then it creates the code that responds to that data.
Slater Victoroff: Yes. I think that's right. I would probably fine-tune it slightly just to say it fine-tunes the code. It's not really creating something from scratch. It's usually doing something in a very, very limited context. You might say, for instance, "I want to multiply the annual contract value of this customer by some number to get the lifetime value." Maybe I don't know exactly what that number is. Maybe I'll use machine learning to figure out exactly what that number is, but in a very, very constrained you're generating a single number form.
Sal Daher: Then you give it all the data and then it comes back to you with the values that you should be adding to it?
Slater Victoroff: Exactly. Even in that case, something as simple as, basically, averaging a bunch of numbers, in some cases, is basically machine learning in a primitive way.
Sal Daher: What sets deep learning apart?
Deep Learning Is Characterized by a Focus on Unstructured Data i.e. Text, Audio and Images
Slater Victoroff: Deep learning really roughly applies specifically to unstructured data. Now unstructured data means text and image and audio. Now, what really sets that apart from the rest of data, and just to maybe clarify, there's unstructured data which is everything I just talked about. There's also structured data and this is stuff that fits in rows and columns where you know exactly what is important ahead of time. That's really the key differentiation because if I ask you based on a structured user record that has credit data, figure out what the credit score should be or not, that's one very particular process.
If you change that instead to say take an image and make some inference based on that, you can see how there's a fundamentally different nature to that process. Before deep learning the way that would happen, and just for an image example, let's say you're trying to tell cats versus dogs. You would have people manually create what was called a feature. These would be incredibly specific. This might be a feature for detecting red circles on green backgrounds and you might write a whole PhD thesis on exactly what problems this was and wasn't useful for.
It'd be incredibly specific. Very, very difficult to build these things. The big revolution of deep learning really is saying this bold thing of we shouldn't make features anymore. People are very bad at this process making features. You should learn the features then as well. Maybe one analogy to tie it to life sciences, because I think it's very close and very helpful, is to draw it back to cochlear implants actually. If you look at very, very early cochlear implants where people just were first tuning into the idea that you could turn these acoustic signals back into something electrical that the body could interpret, the idea went that in order to do this well, you couldn't just send all the signals along.
You had to filter it down in a really, really precise way to just focus on human voices. That was the thought. For decades the way that you made progress and made better cochlear implants was by making these better filters that would do an incrementally better job of focusing the implant and the ear on human voices. It is a very hard problem. It turns out humans speak all over the place and there's a lot of stuff that overlaps with that, so incredibly difficult. The biggest advance they ever made, and this was now many decades ago was they said, what if we just stopped doing that? What if we stopped filtering altogether?
The brain is actually really, really good at making sense of these signals and so what if we just wired in and said let's actually leverage this infrastructure we've already got here to figure out how that should work. It worked wonderfully. It worked dramatically better. I think there's probably some active boosting and things like that but they're not doing filtering to just get human voices anymore in these implants. It's a very, very comparable analogy in deep learning. I believe very much that you had to be very intentional and that you were going to write these very rough statistical rules and things were bags of words.
I didn't think that we would really be able to build up these more comprehensive views of the world. That really is what deep learning gives you. It's an ability to deal with the fuzziness of these problems in a very direct way.
Sal Daher: Okay, what was the problem that you were trying to solve in your dorm room?
Indico’s Original Aim Was to Make Deep Learning Accessible to Software Developers
Slater Victoroff: The problem was fundamentally making deep learning accessible. This is also where I would love to say that I just came up with a beautiful idea for a perfect company in the dorm room and executed on it and here we are, but as all things, I think the path is never so smooth. At first, I was not a deep learning believer at all. Then I was doing these competitions with the person that would go on to become my co-founder who did believe in deep learning and, basically, convinced me as the state of advanced such that deep learning ran away and became so far beyond the traditional techniques I was used to using that I just couldn't keep up.
I said, "Okay, I was wrong, very humbling experience. I want to switch teams." That's really when the crux of Indico's problem came into focus because we realized that while this technology worked really, really well in the lab and you had this unprecedented ability to make sense of text and images and audio that really were very, very inaccessible before. Just incredibly prohibited to do anything with those kinds of data. It was more accessible but the technology was still incredibly difficult to use, wildly impractical in many different ways. The thing that we really focused in on in those early days was this data problem.
Sal Daher: One quick digression, just the name of your co-founder please.
Slater Victoroff: Alec Radford.
Sal Daher: Alec Radford.
Slater Victoroff: After Alec who was the first person I was working with in the dorm room, Diana Yuan and Madison May joined us as co-founders shortly thereafter.
Sal Daher: Excellent. You were trying to figure out a way to make this tremendously powerful technology of deep learning more accessible to people who had unstructured data that they wanted to make sense out of, images, other types of data. Why did you decide to start a company? Why didn't you just be a researcher in artificial intelligence at Olin in the lab there and work on this?
Slater Victoroff Founded Indico Because Most Innovation in Deep learning Was Driven by Industry
Slater Victoroff: We pursued actually several paths independently, I would say. It in fact was not our initial plan to build a company, which is interesting. One of the things though that was really, really shocking that we didn't realize until we started getting very deep is actually at the time the US was actually, basically, non-competitive in the AI space. When you looked at people doing deep learning, there were really only three places in the world where people did this work. It was NYU, University of Toronto and University of Montreal, which are not the usual suspects that you would think.
Then to make matters worse is maybe not the right word, but things changed really dramatically just at the time that we were getting all this figured out, which is that these labs started migrating over to industry. This is when you started seeing DeepMind and Google Brain and Facebook, all being created.
It was a tough decision for us, because what it really meant was there were not really a lot of places for us to go and study this even two years later. We loved the space, frankly, too much to wait. We did start doing stuff at Olin, but, look, there wasn't even a graduate school in the country that really had the kinds of experts that we were looking for outside. Bits and pieces.
There were people definitely doing good work at Stanford and NYU I mentioned, and Berkeley but deep learning is like tatters, very little of it going on in the space. Then we see this massive move over to industry. We just started by spending time with those people. We just said we love this work. We want to do it. We want developers to have access to it. That's really where it started. We made a company because we wanted to try building this stuff out in the real world just to hold a consulting dollar. Then it morphed into what's a product actually that might make this...
Indico’s Experience at Techstars
Sal Daher: Tremendous. 2013 eventually when did you guys end up in Techstars?
Slater Victoroff: We were in the 2014 winter class.
Sal Daher: Right the next year you went to Techstars.
Slater Victoroff: Yes. It was a year in the dorm rooms working from 5:00 PM to 5:00 AM on Sunday nights where really we were an open source project. At the time the most ambitious thing we could possibly imagine was making this technology accessible to ordinary developers. That was really what we were, was APIs and developer tools at the time.
Sal Daher: Right.
Slater Victoroff: After a year we had a really nice pretty natural ground swell of people using and liking the product. We were doing the rounds at hackathons. Even though we frankly were very bad at pitching ourselves at the time, Techstars saw us and took a chance on us at a relatively early stage.
Sal Daher: Who was the head of Techstars at that time?
Slater Victoroff: Semyon Dukach.
Sal Daher: It makes sense. This is purely a Semyon company. Semyon Dukach, One Way Ventures.
Slater Victoroff: He was absolutely great. I feel extremely thankful to have had him as a managing director. He's the kind of guy that every time I remember him, I should be talking to him more than I do.
Sal Daher: Yes.
Semyon Dukach Saw Promise in Indico Others Did Not
Slater Victoroff: He was just an incredible partner of ours and did really help us understand how to navigate a really difficult area. I think he saw a promise in a company that a lot of people didn't necessarily understand at the time.
Sal Daher: The last image that I saw of Semyon was on LinkedIn. He was handing out money to Ukrainian refugees from a charity that he set up. He likes to help. I contributed. I think a lot of people might--
Slater Victoroff: [unintelligible 00:14:47] had that seven K matching going on for it.
Sal Daher: Yes.
Slater Victoroff: Do you still have that link? You could send that to me afterwards.
Sal Daher: I'll send it to you later.
Slater Victoroff: It disappeared in my LinkedIn feed.
Sal Daher: Yes, I think I can dig it up. He's a Russian immigrant to United States and he feels very badly for the Ukrainians who are being made to refugees right now.
Slater Victoroff: Strange for me, because my different sides of the family, actually my grandparents and my great grandparents were from Ukraine. I had some kind of distant family still in Odesa, but it feels similar in some ways and not in others because also at the time where they were living, the question of year by year whether it was technically Ukraine or Russia or Belarus actually, it was hard. Technically, now it would be considered Ukraine.
Sal Daher: Getting back, you went to Techstars. It was by Semyon, say, "Hey, we want these guys here." What did you get out of Techstars? You went into Techstars thinking this, and you came out of Techstars thinking something different.
“...Techstars really helped us understand was that we were real.”
Slater Victoroff: I think the biggest thing that Techstars really helped us understand was that we were real. It doesn't really make sense.
Sal Daher: These guys actually think we're on to something, gosh.
Slater Victoroff: Exactly. I think a lot of entrepreneurs don't necessarily understand just how big that moment is. It clicks over in a very significant way of, "This is a side project to, "Wait a minute, no this is serious, and this is real, and we need to start taking ourselves seriously." Techstars really represented that moment for us. I'll never forget Dip Patel who was at the time the CEO of EcoVent. Because we were working out of the Techstars offices, we were right next to them. We got along really, really well. They were great folks and we were all just tremendous nerds.
The second they heard that we got into the Techstars program, we all rushed over to the corner pub and we did my first and last actually ever Irish car bombs. Which is just such a classic story. I remember he said it was like getting drafted by the NBA. There was a lot of stuff on the softer side that I learned after that through Techstars. A lot of it just understanding, frankly, what to do with a lot of feedback about your company. I think one of the things that was also interesting is that a lot of the lessons from Techstars actually came after the program had ended.
It was a lot of the relationships that we made at the time, a lot of the foundations that were set up. We went into Techstars, knowing we were taking a year off school, but not knowing much else. We came out of Techstars with that oversubscribed $3 million seed round from some of the top VCs on the East Coast. Frankly, 90% of what changed was our mentality and, again, just understanding that just because we were a couple of college students, that alone didn't disqualify us. That if we kept executing like we were and if we kept putting up the numbers that we were, people would believe in us and back us.
Why Slater Victoroff Gave Up the CEO Role
Sal Daher: Tremendous. This is 2013, and then 2017, you switched over to CTO. Tell me about that. That's always interesting. If you care to talk about it.
Slater Victoroff: No, I do. I absolutely do. I think people should talk about it more. I think that's one of the key things. I think a lot of people look at that, it's hard to know exactly what to think. When a CEO moves over, that could mean a whole bunch of different things. The key thing is that I forced our investors to do it. I actually put my foot down and said like, "This is what Indico needs to be successful, and this is what I need to be happy, and they happen to align."
I think the interesting part of the story actually comes from 2014 to 2017, in the time when I was CEO. The thing that is interesting is our space was middling. I think the people that executed with our exact idea, and we executed as well as any of them, but most of them did not end up being particularly successful, well, just because the business model wasn't a particularly good one. We were in that same area and despite those kinds of structural issues. I piloted the ship fine, I would say. I was better than some CEOs and definitely worse than others.
It wasn't the thing that me as the CEO, or even our CTO at the time, no one was doing anything wrong. The thing that was really the shift was twofold. Number one, we realized that we weren't bold enough in our initial framing. We said we have to change the business model. This is not a developer company. We have the people to make a developer company, but that's not what's going to be successful here. We need to shift our people one really big notch over to the left to be enterprise software. That's when I said, "Look, I've been learning for three years. I've learned so much. I've done this, but this is not where I can execute at world-class. We need to get a world-class CEO in here and then I can be a world-class CTO."
Sal Daher: When you say a developer company, you mean a company that develops bespoke software for particular applications and so forth.
Slater Victoroff: No. I don't mean that. I mean the core user. Developing something for developers.
Sal Daher: Rather than just for the enterprise, this is where you're going to scale.
Slater Victoroff: Exactly. When you're making things for developers, your product is code. It's APIs and it's something that a very technical person is going to use. The developer is still important in the enterprise, by the way, but it's one user out of three.
Sal Daher: It's like Steve Jobs and Woz being told by the guy in the computer store, "You can't just sell the circuit. You actually have to have a display and a keyboard and a box because otherwise you're not going to make any money.
Slater Victoroff: That's a great analogy. No, that's absolutely right. I would further the analogy to be we had a lot of really, really great engineers and Woz was in the CEO's spot and we were missing a Steve Jobs and that's what we found in Tom.
Sal Daher: Okay, tell us about your Steve Jobs CEO and how you found that CEO.
Slater Victoroff: It's a funny thing because we did a traditional process. I think one of the things that's important is that at the time we had built up some very real technology and had a well-differentiated market thesis. We had some early enterprise customers and a really good idea of where we wanted to go. I think that's one of the things that's really important. I think of that old quote 'I never want to be a part of a club that would have me as a member.'
Sal Daher: Groucho Marx.
Slater Victoroff: It's a little bit of that mentality where you've got to have something good enough to offer a CEO to be good enough to give back to your company. That was key. We had been at it for several years. We had a highly differentiated technology at the time and we went through a traditional process. We had the money and investors that were behind us that said, yes, it's going to be expensive to go this way, but we're going to go out, we're going to get professionals to find us the best folks that we possibly can.
As always happens in Boston, the second Tom and I were introduced to each other. We've got 10 second-degree connections. My old manager was his old coworker and we just got along very well. One of the things that was always really important in our relationship is that, A, he really cares about the product. I think that's incredibly important for a CEO. He respects the technology very much. Especially for us who are executing in the deep tech side of things that was key.
Sal Daher: Let's give Tom his full name.
Slater Victoroff: Sorry. Thomas R Wilde.
Sal Daher: Thomas R Wilde, CEO of Indico Data.
Slater Victoroff: Yes. He and I worked together for a whole summer to craft one of the early pitch decks that would go on to become Indico 2. He would lead us through the [Series] A and into the really incredible growth that we've seen since then.
Sal Daher: Okay, tell me about the growth.
The Shift to Enterprise Clients and Massive Growth
Slater Victoroff: Yes, absolutely. We shifted over to this enterprise side and it's a flywheel spinning up. Enterprise sales cycles are long, 12 to 18 months. We've had some even longer than that. We had a really clear focus on the idea that we were making this step function more usable. That was the mission. Same idea, same North Star. Even if you look at our very earliest pitch decks, 50% of it still applies. The idea is just that instead of making deep learning these technologies more accessible to developers now we're making them more accessible to non-technical business users.
Sal Daher: People who have need for interpreting data, for understanding data better.
Slater Victoroff: Exactly, the people that are doing it on the front lines today. It actually wasn't until a few years later, but maybe I'll skip forward just because I like this messaging a lot better. When we got our CMO Jeff Thomas on board, this unstructured data platform really started to come into crisp focus, the idea that unstructured really is a capability in its own right. Maybe just to give an example of the kinds of beachhead use cases that you see, document automation is huge today. Thinking about loan applications, thinking about processing invoices, today these are incredibly manual, very expensive, tremendously inconsistent, and error-prone processes.
What we really have concluded is that that's not something that can be done that same way on into the future. There's a need to put real rails around that and it happens to be a really, really good fit for deep learning in terms of being able to plug really assistive technologies into those kinds of workloads.
Sal Daher: Tremendous. Tremendous. Is there anything else that you want to talk about? Your founding experience and then your rapid growth at Indigo Data?
Slater Victoroff: Maybe, I'll toss one more note in there just about the philosophy that we've taken to growth. We think that in a space that is so incredibly hyped like ours is, AI is white-hot, automation is white-hot, document automation in particular is blue-hot even, hotter than that, it's very hard to understand what to focus on as a new company. I think that we in the early days pitched back and forth should we go super verticalized? Should we really try to move down midmarket and maybe bring our ASP down to the 20-30K range? Something along those lines. We made missteps.
Stellar Customer Satisfaction Metrics Was the Secret
We flinched at different times throughout the process. What has really become clear to us through the last couple of years is that the most important thing is really driving that NRR and the CSAT [a metric of customer loyalty] and your success rate in production. What we've found is that it's almost paradoxically--
Sal Daher: Would you unpack NRR?
Slater Victoroff: Yes, I'm sorry. [laughs] NRR is net revenue retention. The idea there is for every dollar I have coming in, in revenue to the start of the year, how many dollars are those existing customers getting me at the end of the year? It's net of churn but the idea really is that if you're delivering value inside the enterprise and you've got a good SaaS model that scales with usage in some way, it's not that it has to be pure consumption-based, pure PLG, but there's some scaling.
Sal Daher: Okay, it's an average of the ARR for each customer or is it all the revenue and all the customers you have and so you're growing usage in some, you're losing in others but if it's growing overall?
Slater Victoroff: It's importantly of the existing ones coming into the year. It doesn't include signing new customers. That gets counted separately. Some can balance out. For instance, if I have customer A churn but customer B doubles their license, that comes out in the wash and then that would be 100% NRR.
Sal Daher: Okay, so it's like in retail same-store sales.
Slater Victoroff: Yes, that's a great analogy. For SaaS companies, for a benchmark, 125% is a really good performing company. That's very, very high CSAT and it's even very common in e-commerce or something to have an NRR that's bellow 100% where churn is high but you got new logos coming in that still lead to growth.
Sal Daher: Right.
Slater Victoroff: We actually had 149% which is great on two sides. I think, number one, it makes it very clear to everyone in the space that we're delivering successful production, these customers like us, they're buying more, and it's a fantastic litmus test because most people in our space have had the inverse happen. They've had a lot of churn, they had negative renewals where they were losing dollars into the next year. They've had ASPs down significantly below what their targets are. We've actually not had that.
Sal Daher: ASP is?
Slater Victoroff: Sorry, average selling price.
Sal Daher: You have this NRR that was 1.49. That means a customer finds you and they can't let go of you. They want more and more of you.
Slater Victoroff: Exactly. That's exactly right. It means that they're upselling themselves 50% on average year over year.
Sal Daher: You reach escape velocity pretty fast going like that.
Slater Victoroff: It certainly helps. It's one of the things where that's not going away. That turns into a real sustainable base and it just gives you breathing room to figure out the rest of your business. There's nothing worse than trying to pour gasoline on a fire while you're heading in the wrong direction.
Sal Daher: Yes.
Slater Victoroff: Having that really, really heavy investment in CSAT lets us--
Sal Daher: It doesn't work to pour gasoline on a fire when you're taking on water.
Slater Victoroff: Yes, exactly. That's exactly right. It really gave us the space, emphasizing CSAT really heavily on to be more diligent about how we build our sales organization and really do that right because it means that we didn't have to rely as much on the new logos to drive that growth.
Slater Victoroff Was an Unlikely Founder
Sal Daher: Okay, all right, so let's turn a little bit to this idea of becoming an entrepreneur. I know you became an entrepreneur because there wasn't anywhere else you could do what you were doing but did you have an entrepreneurial model, a mentor in your life? Were any of your parents entrepreneurs?
Slater Victoroff: No. When I showed up at college, if you told me that I was going to be an entrepreneur, I would not have believed you. I would have laughed you out of the room, saying, "No way, no how. You must not know me." No, my parents were lawyers. My extended family are doctors. I think that there was maybe a great uncle twice removed who was an entrepreneur for a brief period of time, but I never met him. No, there was no role model in that kind of way.
Sal Daher: It was just, basically, you were wanting to pursue this technology. You were excited about it and the only place you could do it was in industry because all these people were going off to industry. They were being absconded from the universities. That wasn't happening, it was happening in industry. I'm an investor, for example, in a company that's using machine learning for image recognition. It's hard hiring machine learning people. What's a hint? Give me a hint. If you're a startup trying to hire a machine learning person, do you have any pro tips?
About Hiring Machine Learning Talent
Slater Victoroff: My pro tip actually is you probably shouldn't be hiring a machine learning person.
[crosstalk]
That's the secret.
Sal Daher: You should sign up with Indico Data.
Slater Victoroff: No, no, no, no, I don't actually mean it that way. Most people aren't asking this at a Fortune 500 company that's going to be an Indico customer. This is really for the startups. We want to get after ML. I will say deep learning is significantly harder than traditional ML. I will say that doing traditional ML, if you've got structured data and you're just trying to do some boosting or traditional models on that, you can do that with a few months of Kaggle education. You don't need to get a super sophisticated PhD to do. By the way, that's 80% of the ML that's out there.
Sal Daher: This is completely unstructured data.
Slater Victoroff: Right, and that's where things get more interesting. Then I would even break that down further because, believe it or not, computer vision is actually by far the easiest kind of unstructured data from a talent perspective. Even then, what I generally say is that focus on data architecture first. I say, usually, first get a data architect, if you're trying to hire an ML researcher. Until you've got a data architect, even if you've got the best ML talent in the world, they're not going to be effective. On the flip side, having great data architecture without an ML researcher actually adds tremendous value.
Sal Daher: It does add value.
Slater Victoroff: Tremendous value. Yes. Then from there, even the second person you should hire isn't even an ML researcher. The second person, you should hire is someone with a really crisp product mind that is more data-oriented. That's the other thing. Without both the foundation to build worthwhile ML and the product strategy to understand how it's going to fit into your broader business, it's not going to be valuable. I think that's the biggest mistake that a lot of people make, is if you're trying to hire an ML researcher before you've got the whole foundation set up, and if you've got the mentality that I'm just going to hire an ML researcher and give this all to them, you're not setting them up for success.
It's going to be really hard to find people that really check that box. Then on the flip side, once you've got all those set up, actually large swaths of talent that might not have been as valuable before are suddenly accessible. You can get someone who's coming out of a master's program, and they're going to be able to add tremendous value. You really don't need a PhD for the vast majority of things you're going to be doing.
Sal Daher: Great. Slater, let's talk a little bit about what you're doing at 406 Ventures as an entrepreneur in residence, and what things 406 is looking for. 406 Ventures is a very forward-looking very thoughtful venture capital fund here in Boston.
Slater Victoroff: Yes, I love working with the 406 folks. Obviously, they're a backer of Indico. That's how I got connected with them the first time around. Really, what I'm interested in 406's role, as all of my roles generally advising in angel investing, is deep tech with a focus on AI. My view very much is that, and it's probably because I'm an engineer at heart this just happens to be the way that I look at the world, there has been this really significant breakthrough technology that gives us a whole set of tools that we didn't have yesterday. We're still figuring out exactly what to do with those tools.
I really, really like companies that both understand the new tools that have come out. There are still plenty of places where we're using 30-year-old technologies still adding value, but that's just not what I'm personally interested in. I am really interested in these places where you've got more the bleeding edge tech that is this large-scale language modeling and reinforcement learning and these deep learning approaches, and really using that in a meaningful way to execute on some meaningful use case in a way that you couldn't yesterday. I really like especially when people dovetail the technology with a meaningfully improved user experience that they can deliver, and they even present things fundamentally from that lens, but then it's a by the way, you can only do this now with this new technology.
Slater’s Work at .406 Ventures
Sal Daher: You spend an afternoon a week or something at 406 listening to pitches? How are you involved with them?
Slater Victoroff: Primarily, I'll be doing technical diligence. They'll be considering an investment in a company. I get called in to sniff around the AI. I will say, the .406 folks to their credit have gotten very good at doing that on their own, so now I get called in in those gray area cases more than anything. From time to time, .406 will also be crafting theses. Looking at broad swathes of companies and companies trying to figure out, "I've seen six pitches in this space today. Is there something here? Is this fluff or not?" Primarily it's technical diligence, and then we also do some special projects from time to time.
EVQLV
Sal Daher: Let's talk about you're an advisor at EVQLV (pronounced “evolve”). Tell us what they're doing and what attracts you to the company.
Slater Victoroff: Absolutely.
Sal Daher: That's a company that I think they're funded by TBD angels where we're both members. I think the guys at Walmart also invested and so forth. Anyway, tell us what they're doing and tell us what you find interesting.
Slater Victoroff: It's exactly what I was just talking about, an area where you've got an intersection between this really bleeding edge AI and a use case really that's going to impact a lot of lives. There's this very interesting parallel, it turns out, between these modern large scale language comprehension models. We do a lot of that, obviously, in the document space. There's an analogy between that and protein modeling.
Folks have heard of AlphaFold. This is a great example where you can go read up more about this. It's a very, very cool space where it's a set of tools now to think about and constrain protein modeling that you never had before. While I will say that I'm not an expert in bio in any way, shape or form, unfortunately, I wish I was smart enough for that, really what I'm helping them understand is we've got this amazing new set of tools. How can we really use those effectively? What are these useful for? What are these not useful for? Learning enough about drug discovery and the process to be dangerous.
Sal Daher: Right, that's good. How did you connect with EVQLV?
Slater Victoroff: They came through Techstars.
Sal Daher: Those are Techstars, okay.
Slater Victoroff: Yes. I do a lot of advising at Techstars still. Unfortunately, right now I'm a little busy, so I had to take a break for this last class, but I'm hoping to get back into it.
What Sets Olin College Apart
Sal Daher: Excellent. Well, let's do this. Let's talk a little bit about Olin, and what Olin is because a lot of people going to say, "Olin I never heard of Olin". Olin College. I like to say Olin College is like MIT, but where everybody knows each other. An undergrad class at MIT is 1,000 people. Larry Summers was in my undergrad class at MIT. I never ran across Larry Summers. Tell us the experience at Olin where your class is 80 people.
Slater Victoroff: One of the things that's actually interesting, I'm one of the only people probably where going to Olin was actually an increase in class size for me. My graduating class in high school actually had 42 students in it. Olin was twice as big, 85 kids. Amazing.
Sal Daher: Well, give a plug to your high school.
Slater Victoroff: I went to public school, if you can believe it. I went to North Hollywood High School and there is a special small magnet program there that I attended with some real incredible teachers. It is very small. It was caught in a very weird no man's land, where there's a test you have to take to get in, and then the district stopped administering the test, which is why the classes got so small. I don't know. Really, I loved it. A teacher there in particular Altaire Maine, I would not have gotten through high school without him, frankly.
Sal Daher: That is tremendous, having an inspiring mentor like that. It's tremendous in the sense that it gives you that direction that the mentor gives you, and it also teaches you the value of being a mentor for when you are in a position to do that later on in life. My dad was a mathematician and he thought that everybody could be a mathematician if they found the right mentor into mathematics. If you're not a mathematician, it's because you didn't find the right person to lead you into mathematics. [laughs] His working assumption.
Slater Victoroff: I totally agree with that. I think that it's actually one of the great tragedies, frankly, that people have got such negative associations with math. I think a lot of it has to do with the inverse of that. You had a crummy teacher. Math was hard for them, so gosh darn, it's going to be hard for you too. I don't know, I think it's awful. I think that math is taught horribly. I think that it's busy work. I self-educated myself. The biggest breakthrough I made in math was when I stopped listening in lectures and I just started reading the textbook. That really says something about the state of math education.
How Textbooks Empower the Self-Directed Learners
Sal Daher: You read a textbook?
Slater Victoroff: I read a textbook. Imagine it. It turns out those things are useful and no one ever tells you that.
Sal Daher: I was always wondering why my daughters had textbooks, because they never read them. They relied entirely on the lectures. "Why do you have textbooks?" "Nobody reads the textbooks, dad." They were excellent students, but it was all like [unintelligible 00:40:54].
Slater Victoroff: It's not how you do school actually.
Sal Daher: I couldn't have gotten through just with the lectures. I had to read the textbooks.
Slater Victoroff: It's interesting. I went almost the opposite way. I think, unfortunately this probably happens to more kids than not. I didn't have a huge amount of money and textbooks are insanely expensive. What I used to do is I would go down to the thrift store, and every thrift store has a book section. That's where people are just like getting rid of everything they've got and so there are these really high-quality college textbooks in there. It's 1 college textbook in 300 cookbooks and romance novels and other stuff. You have to look really carefully to find them. That's what I used to do. I'd spend a whole afternoon there just going through every book and then once you find them, they'll give them to you for a quarter. Maybe not anymore, but that's how it used run.
Sal Daher: It's a funny thing. When I was in high school and college, textbooks were much smaller than they are right now because graphics were so expensive. The spreadsheets are expensive and all the stuff. Now they're so heavy and nobody reads them. In the old days they were smaller. Now they're huge and high school students can't carry them. They have to split them up.
Slater Victoroff: It's funny. This maybe goes a little bit to my philosophy and then I'll use this to come back around to Olin. I had tons of textbooks. I brought tons of textbooks back and forth to school every day. Also, starting from my senior year of high school on through the entirety of college, I never used a single school supplied textbook. Never once did I buy a single one. I'm not into this. I love textbooks. I'm just like, "This is a racket." As an entrepreneur I'm surprisingly extremely stubborn. I just object. I'm not going to participate in this racket. I love textbooks. I think there's a tremendous amount of value to be had there. I just think that the ones that are required for classes happen to be too expensive for any person to actually buy.
Sal Daher: Tremendous. Anyway, your experience with Olin. 80 students per class, MIT types.
Slater Victoroff: Let me maybe tweak around, because I think here's an interesting piece. I think the thing about Olin that is really distinctive, not that the small class sizes aren't important, if people come away with one idea from Olin, I want them to remember project-based education. This is the really revolutionary idea of Olin, is it's almost Montessori-style education for engineering.
As a traditional, very hardcore or textbook oriented math and science student in high school, I'm like, "That sounds fluffy as hell." The way that I thought about it is I had completely separated coursework and learning. I was like school is something that takes away from your time to learn and I assumed that I was going to be one of those never go to class, show up for the final, get an A in the course kind of kids. Olin was the first place that actually convinced me that I could do classwork and learn at the same time.
Sal Daher: What a discovery.
Slater Victoroff: This is the power of project-based education, is you've got a tremendous amount of flexibility to choose what you're going to work on. You pick something that actually is going to be challenging enough to keep you interested in the class. Always, pretty much, you're going to fail because it's so hard to build things, but the thing is you grind yourself to such a fine point trying to get the thing to work, you learn way more than you ever would if you were just trying to get an A on the final.
Why Slater Victoroff Does Not Invest in the Life Sciences
Sal Daher: Tremendous. Slater, in the interest of not keeping you too long, two things. I'm going to ask you why you're not an investor, I know that you invest, why you're not an investor in life science companies. That's my idée fixe. Then we'll close out with your closing statement what you want to communicate to the listeners. Why, Slater, are you not investing in early-stage web lab life science companies that I'm so excited about?
Slater Victoroff: I'll put the asterisk on it, obviously, that any sufficiently interesting company an investor will take a look at but maybe I'll give two half-answers to that with that asterisk in mind. One is that I'm just, frankly, not smart enough. I think that there's fields that I understand and as a technical guy I really like to understand all of the ins and outs and biology is really hard. I've got a lot of smart friends that do really understand it, which I'm very thankful for, and they're just smarter than I am. It clicks with them in a way that doesn't quite.
I have trouble sifting through it just for that reason. That leads me to secondary place. I am very, very interested in life sciences companies where it intersects with technology where I really feel like I do understand that, and really importantly, I can add value. Executing on deep tech in the AI space has a really unique and profound set of challenges. That's really what in the investing and advising game for is if I don't feel like I can speak to your experience and help you avoid some issue, I don't want to just be adding more noise into your life. You've got enough of that as an entrepreneur.
Sal Daher: I appreciate that. The reason I'm so obsessed is that it's a space where there's a lot happening, and there's just not enough investors. Precisely one of the reasons is that it intimidates people. I'm sure it's not that it intimidates you, but it's just that you have more--
Slater Victoroff: No, it does. It intimidates me. I'll admit that openly.
Sal Daher: You could pick it up very quickly. The thing is that you're always starting at ground zero. What I find with life science companies is that I'm always starting at ground zero. Having invested in 15, 20 different life science companies doesn't buy you a lot. Maybe I understand it may be a little bit of a tempo in a couple of things but when you get down to the technology, it could be you're a few molecules away, you might as well be 10.
Slater Victoroff: I think that is the thing. I think everyone's got a different level of how familiar do they have to be? How deeply do they have to understand for them to really feel comfortable? I'll say this as someone who dropped out from college, but I still do generically believe in education. I just think that's because Olin did too good at the job and they taught me whatever, college and a half let's talk in three years.
Sal Daher: You had your own project. Indico.
Slater Victoroff: Exactly but I really don't feel that way about biology. There are definitely spaces where I feel like I can get up the path quickly but that is the thing. I think that the reason why bio is still very heavily dominated by PhDs is you need a lab to play around in really to get a serious expertise.
Sal Daher: It requires almost a monastic dedication in the science before you can make any contributions. That’s the thing. There are no shortcuts in the life sciences.
Slater Victoroff: No, there aren't. Software, I can literally just cram time into that, and I get expertise per unit time. Not the case in the life sciences. It doesn't matter how long I stay in the lab. It's going to take six months for the culture to grow.
Sal Daher: That's very true. Anyway, parting thoughts?
Parting Thoughts from Slater Victoroff
Slater Victoroff: Boston's a great place for startups. People should make more companies. Whether they're in the life sciences or AI I think there's a lot of space in both. I think there's a lot of tremendous promise here. Frankly, I wish that I saw more companies that draw both together to a fine point because I think we've got great DNA for all of that here.
Sal Daher: Ah, Boston has a tremendous environment for starting companies.
Slater Victoroff: It really helped me.
Sal Daher: It has so much to offer. Despite the reputation of New Englanders as being standoffish and so forth, it's tremendously collaborative. It's not competitive. It certainly has its drawbacks, but I find the ecosystem here to be tremendously supportive. I can always reach out to people. They get back to you. It's awesome. It really is awesome.
Slater Victoroff: I totally agree. I'll say probably all that is deserved. I think definitely New Englanders maybe there's a low tolerance for BS. My friend says it's because it's cold all the time. Everyone's always trying to get to the point.
Sal Daher: They're stamping their feet because they want to get inside. It's cold. Get to the point because it's cold.
Slater Victoroff: I think that once you start looking through that lens at it, it starts to make a lot more sense. I think the thing that I do really love about Boston is that combination small town, big city, where very vibrant, tremendous amount of companies. Just in this conversation, we've probably talked about half a dozen companies I hadn't even heard before the call. They're probably all amazing. At the same time, everyone knows each other. We probably also found half a dozen people between the two that we're both great friends with and that's Boston in a nutshell right there.
Sal Daher: It's a miracle the geography and the culture here that that happens. It's both a big town and a little town. It's much bigger than San Francisco, and the Bay Area, in the sense that-
Slater Victoroff: It's much bigger.
Sal Daher: -there are other industries. There's a lot going on here other than tech. At the same time, it's very intimate because of the geography and the layout. The subway system, you run into people. It's like walking down Mass Ave they're running. I see Armon Sharei going into a café, with his wife and his baby, a founder of one of the startups that I invested early on. It's now public. That's the great thing about the Boston area, Cambridge, and Boston in general.
Slater Victoroff: I couldn't agree more.
Sal Daher: Where are you, by the way?
Slater Victoroff: Sorry. I'm in East Somerville right now.
Sal Daher: Just a stone's throw from here.
Slater Victoroff: I think, it's actually one of the things that has kept me in Boston is probably one of the most durable things is it really is a walking town. You don't have to own a car and I think that's such a bizarre and foreign thing to folks in the US. It was crazy to me coming from LA which is not a city you can walk and get somewhere. It just makes things so pleasant. You get to run into people all the time. It's a load off your mind constantly and it gives it just a totally different feel.
Sal Daher: It's a great town. It's tremendous. Slater Victoroff, founder of Indico Data in his dorm room at Olin College, I thank you very much for being on the Angel Invest Boston Podcast.
Slater Victoroff: Thanks for having me.
Sal Daher: This is awesome. This is Angel Invest Boston. I'm Sal Daher. Thanks for listening.
[music]
Sal Daher: I'm glad you were able to join us. Our engineer is Raul Rosa. Our theme was composed by John McKusick. Our graphic design by Katharine Woodman-Maynard. Our host is coached by Grace Daher.