The below is a full (unedited), machine-generated transcript of a Youtube session / podcasting episode I recorded with Lee Edwards, partner at Root Venturs in Q4 2019. You can view the video/listen to the podcast on Youtube, Apple Podcast, Stitcher or wherever you get your podcasts.
Trailer
Erasmus Elsner 0:01
Hi, and welcome to another episode of Sand Hill Road, the show where I talk to successful startup founders and venture capitalists about the companies that they build and invest in. The goal, like always, is to give you a sense of what it’s like to be in their shoes, how their businesses tick, how they got to where they are and to learn from their many successes and mistakes. And today, I’m thrilled to be joined by Lee Edwards, partner at Root Ventures. Lee has a diverse, both technical and operational background, which I think uniquely qualifies him for his role at Root Ventures. Before joining Root last year, he most recently was the CTO of Teespring. And before that, he held positions both as a mechanical and as a software engineer at startups, such as Pivotal Labs and iRobot. Root Ventures is a San Francisco based frontier tech venture firm, which closed its first $30 million fund back in 2015. And it closed its second fund last year in 2018, with LP commitments totaling exactly $76.7269 million. Why this exact number you may ask? Well, because Root Ventures is composed of real engineering nerds, and 767.269 miles per hour is the speed of sound in dry air at 20 degrees Celsius. And it is exactly this kind of engineering mindset and attitude that sets the tone at Root Ventures. And it’s also the reason why Lee decided to join the firm last year. But let’s hear it from Lee himself and let’s jump right in.
Interview
All right, I’m super excited to be joined today by Lee Edwards, who is a partner at Root Ventures. And before we begin, I want to give Lee a lot of credit for being one of the first people to agree to be on my show way before I did my first recording, I’ve done a bunch of them now. But he without seeing any of my content, he agreed right away. So he was really an early backer of the show before it was a show. So a lot of credit to you. And before I want to dive into the work that you do at Root Ventures, I want to take a step back and really find out who is leap. So you have a bit of a unusual background in that you are both a software engineer and a mechanical engineer. I think when you first graduated from Olin college with a degree in systems engineering, you actually took a role as a mechanical engineer at iRobot. And it was only later that you basically retrained as a software engineer, and held various positions and startups like Pivotal Labs site tour, and Groupon life. And most recently, you were the CTO of Teespring, which is a Series B stage startup in San Francisco, in the apparel industry, and which operates at the intersection of software and hardware. And I think if I met you two years ago, I would have said that you’re the perfect technical co-founder for a startup. As you operated really at this intersection of software and hardware. But instead, you decided to go from the operator role into a venture capital role. So take us back to that moment. And I mean, I know that you were an angel investor in companies such as Skydio, Lever and Enzyme before that. But what really was your decision process when you decided to join Root Ventures.
Lee Edwards 3:28
So I had been angel investing in some friends companies, and that’s where some of those companies you mentioned, came about. So it’s something I was kind of doing opportunistically, which I think happens to a lot of people out here in the Valley when your friends start companies and you want to support them, or maybe you have a set of expertise that they want to bring you in on the cap table. So I was doing that. And Bloomberg Beta, which is Bloomberg LPs corporate venture arm, actually offered to start backing my investments by putting me into their scout program. So I started basically doing that to just bring more capital into companies that were already in my network. But when I started kind of spending nights and weekends on it, and trying to source companies and finding companies like, like Dispea stories, which is doing audio romance novels for women, invested in Daisy, which is a London based company, founded by Maisie Williams that’s working on, essentially, how to how to disintermediate agents and folks out of Hollywood and let the creatives kind of get closer to their audience. I found that I really enjoyed that. And so I felt if I had more time, I could do more of it, and I could do a better job of it. So Teespring was in a good place at that point about January of last year, and there’s stuff in the press, you can read about it. We went through a lot of tough times what we definitely pulled out on the other side. So I felt comfortable leaving and kind of focusing on investing full time. My original plan was to kind of do this on my own, maybe start my own small fund. And I invested in a company called Esper as an angel that is much closer to the area that I’m investing in. Now, they’re sort of enterprise operating system and development tools set for essentially Android devices, touchscreen devices. So you can think about touchscreens on refrigerators or on cars, or even not touchscreen devices, like Amazon, Alexa actually runs on Android. And so what’s the flavor of Android is not for consumers, it’s actually for enterprise instead of cloud services and tools that you need to, to actually have an effective development process. So invested in that company actually committed very early, I think it might have been the first commitment. And then Root Ventures ended up leading around, and I’ll be done, the general partner at root called me in and was trying to figure out if I was a worthwhile angel to have on the cap table, or if he should throw me out. And so I wouldn’t, I wouldn’t let him throw me out. And we ended up kind of talking over a period of time and decided to try to try out the relationship as a venture partner. So I did that basically August through December of last year. And then by January decided I wanted to join full time. So they have a nice advantage of I think, moving from operator to venture, not everyone likes it, I know a lot of people that have started to do it. And then they decide to leave venture pretty quickly when they realize that they’re much more interested in kind of hands on building. And I had a year to basically spend doing that myself. And you know, being a being an investor in a fund is very, very different from being an angel investor in many ways, but at least had the opportunity to figure out like, do I want to work with founders in this way, versus, you know, being an engineer or manager or even an executive or founder at a startup. So there’s no shock to me that, you know, the past year of being a partner at root. Now, that’s interesting from almost being kicked off the cap table to joining the leader of the round right.
Erasmus Elsner 6:57
Now, let’s talk a little bit about Root Ventures. In the great scheme of things, it’s still a pretty young firm, which raised its first fund in 2015. And it has this really unique investment thesis, I think, on hard tech, which I think you’ve talked about this somewhere else. If you think about the history of venture capital, it is meant to finance “what would not be financed otherwise”. And in that sense, I really like this thesis of Root Ventures, which is trying to do exactly that. And maybe let’s start with the name. So Root Ventures is a reference to the pseudo or super user in the Unix computer system. So what does this mean concretely, for founders, the founders have to give you root access once you invest? Or that you get really into the weeds of the software architecture? What is the message to founders there? And what is root Venture all about?
Lee Edwards 7:56
Yeah, well, we definitely don’t require super user access, I think we’re happy with a managed account and letting the founders run their own root account. But like that you caught the reference, not everyone does. The obvious. The obvious other thing right is the roots of the tree. So if you see our logo, and it’s actually silver traces on a black PCB, but it’s in the shape of a tree, we were all come from an engineering background. And I think actually all of us have touched at least touched a little bit of a few different disciplines of engineering. So I think the founders that we work with, get the joke know what root is, although what we have found is that Australian founders have a different definition for what root means, which is vulgar. That’s not something I think I’ll be done thought about when you pick the name. So the firm is about about six years old, we’re on fun too. And the first seven that we call sort of fun zero, the set six SPV is that we invest that I’ll be done sort of raised and invested directly a lot of interference kind of get this start this way. If the fund managers a first time manager, they’re often doing angel investments, bringing in people to back them kind of like the path that I was going down to some of those investments included particle which is internet connected microcontroller, which just announced a large series C. And creator, which is a hamburger making robot that is doing quite well. COSLA Khosla ventures has invested a lot since then basically, the idea has always been technical founders technical problems. What’s interesting is we’re we tend to be known for doing hardware because we do hardware and I think most firms shy away from as you were saying, I think that some capital in Silicon Valley is kind of averse to technical risk, and maybe that’s okay. I mean, if the if the partners that those firms don’t have that background, and they feel like they’re better aligned towards understanding risk, and helping founders in other spaces, like commercialization, go to market, that kind of stuff. That’s great. They should stay out of deep tech, we like hard problems of all kinds, you know, even back to the name root, it’s a computer science term. So we actually have a number of companies that are pure software and that’s that’s really I’m going to be focusing,
Erasmus Elsner 10:01
Let’s talk a little bit about your personal investment sweetspot. Now, and I saw on your website that you are focusing on heart software, I’m quoting from your website here, tools and services for engineers and data scientists, software, infrastructure, AI, machine learning, computer vision, and AR and VR. So I mean, this sounds really daunting, and it surely scares away all the dating apps and the productivity apps. But it’s still it’s still like a huge space. So maybe you can narrow it down a little bit what your personal dealflow funnel looks a little bit like,
Lee Edwards 10:34
Yeah, it’s interesting that that definitely is the case, when you kind of put that out there. I think a big part of the venture firm on the newer side is trying to build something that is a clear message that founders will come to you when they’re in the space or that other investors will think of you when they’re looking for someone to lead or co-lead in one of their investments. So yeah, I mean, I’m interested the past couple investments I’ve done can’t quite talk about yet. But both are in what I would call Well, one is in what I would call sort of data science enablement, or AI enablement tooling for building a specific set of data science tools and models. And the other one is more on the data engineering side solving. One of the biggest problems that exists in data engineering that most engineers think can’t really be solved, excited about that kind of stuff. I mean, we were essentially doing two to three investments per year, I’ve been at the firm for a year. So stuff that I expect to be doing and looking at in the future. definitely interested in other kinds of applied AI. So a couple companies in our portfolio that are kind of like this already. instrumental is a company that’s doing computer vision that sits at the end of the factory line. So products that come up with the line that can detect anomalies, or defects. And what’s really amazing about what they’re doing with computer vision and anomaly detection, is that they can train, they can train a model on 20 images. So it can be very useful for the kind of factory line where products are changing throughout the day. And so it’s really useful for say, like MPI, like new product introduction, or kind of like prototype level manufacturing, where the quality systems are pretty early with another company called her to go that’s using computer vision for robots that detect and pick strawberries out of the farm. So that’s that I would call a hardware company instrumental is I you know, the other thing is, I think a lot of times people think that something’s a hardware company, if it involves hardware, but I tend to think of instrumental as a software company, they don’t, they didn’t have any mechanical engineers, when they got started the sort of cameras off the shelf, the fixture that holds the camera was off the shelf, and certainly down the line, though, you know, start building a little bit more of that, but it’s really driven by software. And the same is true of Particle, it’s, uh, it happens to provide a microcontroller that’s connected to the internet. And when they got started, they had sort of the highest performing lowest cost one. But what Zack Cipolla, the CEO did that was really smart, was he anticipated, anytime you make something like that, it’s going to get copied in this in this environment, this global economy, there’s there really isn’t any proprietary PCB. So sure enough, yeah, there’s lots of stuff out there in China, that’s lower as low cost or lower cost. But what particles really selling where the where they’re making their money is the software services built around it. So another company called anthropology that’s doing generative CAD design. So they’re pure software for mechanical engineers, that are doing advanced materials, manufacturing, and advanced geometry, trees in ways that traditional CAD applications like SolidWorks and pro engineer, Autodesk, don’t do a good job of they’re not they’re not designed for this, they’re sort of you can think of them as essentially declarative modeling, whereas, and topology is kind of constraints based. So solve this equation, minimize the weight of the solid, given this, you know, set of thermal properties and convection equation. So what it’s doing is actually pretty, pretty insane. And the customers are amazing. We’re not allowed to talk about most of them. They’re doing really, really good stuff.
Erasmus Elsner 14:00
Very interesting. So let’s talk about another segment that you’ve been looking at, which is the commercial open source segment. I want to kick off the discussion with a tweet of yours that I really like. And so basically, you’re quoting there, someone’s saying that Silicon Valley profits off of open source and doesn’t contribute back. And then you basically wrote that I don’t know how much value you think React, TensorFlow, Kubernetes, Airflow, Golang have added to the OSS community, but there’s bigger impact to be made than closing issues on other people’s projects. I really like this tweet, because we’ve had a lot of discussion about value contribution and open source, obviously, with cloud providers moving quite heavily into this space with the whole Redis and AWS discussion, the whole Commons class discussion. We have the announcement today from Dockers, basically selling their enterprise solution and their show got stolen from Kubernetes, once Google open sourced Kubernetes right. Your tweet brought to the fore something new that I haven’t heard in that in that sense. And that’s really the geo-local contribution level. And if I can play devil’s advocate here for a second, and push back a little bit on your tweet, one could say that open sources is this genius way of Silicon Valley to basically crowdsource a lot of the infrastructure codebase across the world, but still retain much of the monetization layer in Silicon Valley. And thinking of projects like React, obviously, it’s used in other projects, but a lot of it is still monetized through Facebook, or think about things like Kafka, which came out of LinkedIn, and is still heavily monetized by LinkedIn. Obviously, they’re also doing a lot of contributions. My argument here being that there’s a geo local dimension to this whole value contribution. So talk to me a little bit about how you how you view that.
Lee Edwards 15:53
Yeah, I mean, I think that there is it’s certainly the case that open source was sort of started, like, well, outside of Silicon Valley. I mean, I think it was much more distributed teams, if you look at ngannou, and things like that, what I think has been interesting to see is the kinds of innovation that’s coming out of places like Google, kind of less so Apple to some extent, although I mean, well, actually, that’s an interesting one to see. Right. And then moving from Objective C to Swift, and open sourcing Swift. Really interesting trend, right? Microsoft moving from proprietary in the 90s. Right, that was, that was the whole deal is proprietary everything. And so now, well, Microsoft owns GitHub. It’s also been open sourcing a lot of their libraries and frameworks and working a lot better with the community. So these companies could hang on to this IP, you could imagine Google saying, AI is so important. Why would we release TensorFlow? It’s just kind of getting rid of our competitive advantage. So I think that there, you might look at that and say, Hey, well, it seems like Google is crowdsourcing their development. But I don’t view it that way. I think partly because I don’t think that really works. I’ve been in companies where, you know, people who are not engineers have sort of suggested, hey, we could get a bunch of free labor if we open sourced this, but it doesn’t work that way. Because you actually have to build the community. Nobody just sort of finds a project and starts contributing to it in a meaningful way. If you look at the most successful OSS projects that are out there, have amazing communities. And when they are things like go Lang that have come out of companies, the company is spending considerable resources to do the sort of developer marketing, running the conferences, paying people full time to do this. And you’ve seen people like Guido from Python inventor working at Microsoft, you see Matt, who invented Ruby working at Heroku, Salesforce. These companies are putting money into it. And if you think about the value that those projects that I listed have created in the in the community, it’s it’s enormous. And I think if anything, the thing to worry about is like yeah, did did Kubernetes take a lot of value away from Docker? Yeah, possibly, who’s to say where Docker goes from here, right, divesting the enterprise business, but certainly investors and employees, a Docker might sort of be upset about the existence of Kubernetes being free and available for everyone. But for the rest of us building software in the world, I think it’s kind of amazing that Kubernetes is free. Like a lot of these projects, when you look at Kubernetes, right, I can’t tell the number of times that we’ve basically built a buggy, more feature barren version of Kubernetes. Before that came out is like enormous, almost every company was building like a crappy Kubernetes. And certainly, that’s the case with all these JavaScript frameworks and react, right? How many times were we reinventing the JavaScript framework, and now it feels like Fingers crossed, like react may actually finally have like kind of the foothold into staying power in the space, I guess there’s kind of a knee jerk reaction to sort of dislike large corporations. Another interesting trend that VCs will often talk about is the sort of shift from innovation in concentrated organizations. Like if you look at Bell Labs, if you look at where a lot of technology innovation was happening, 50ies 60ies 70ies is really like, you know, post war up until kind of the Dotcom, first Dotcom era, innovation was happening out of large companies. And then we now we’ve seen a lot of innovation coming out of startups. And I think we’re starting to see, in some cases, a lot of innovation coming into large companies. Again, that’s actually been happening for a while. But what’s interesting, that’s different from what Bell Labs was doing is a lot of this innovation is getting sent out to the world for free. And so I’m not sure that it makes sense to invest in golang as a language or TensorFlow as a framework. But it’s interesting to look at the world and say, What can people unlock now that they have access to these technologies for free? I don’t know the specific answers to those questions, or I’d be starting those companies. But I’m always looking for founders that are maybe they’re a domain expert in something so like, instrumental who I mentioned and said, let’s see, the founder is an expert at supply chain, she was at Apple, the availability of cheap, easy free computer vision models that you can build that that compiled that you can retrain Didn’t she didn’t have to build any of that from scratch. She just needed to be a domain expert in supply chain. That’s what I’m looking for out there. I think that as that bar of what what is easy for people to implement, what do they sort of get for free that unlocks so much potential for people that have to know that have less than they would have had to know before.
Erasmus Elsner 20:01
So let’s talk a little bit about the different business models in open source. And the three main business models are the “services” business model, pioneered by RedHat, which lacks a little bit in scalability. But often times open source companies start out with this, then there’s obviously the “open core” model. And then there’s the Saas model. How do you think about the models in terms of defensibility? And also in terms of scalability?
Lee Edwards 20:25
I think that they kind of have different applications, they have sort of different pros and cons, depending on the application. So I think the services model is an interesting one where even though you see companies like you mentioned Red Hat as a service company, there is the card for the double digit billions kind of insane. That was a services business. And I think traditionally, people think of services as not venture bankable because it’s, you know, scales linearly with labor. But when I think of services for these kinds of businesses, I think of services as lead gen as ways to bring on customers and kind of understand what parts of the roadmap are important to kind of CO develop your application, and its contract expansion. So I did a, I did a teardown of the pivotal s one when they IPO in April of last year. And I’m not an expert in any of that. So I saw that. So one thing that you can kind of pull out of that data is that pivotal is using services to their landing contracts. They’re getting these large customers to use the product, Pivotal Cloud Foundry and related products. They’ve got these companies paying for their staff and their cloud SaaS offering, which is their extremely high margin part of the business. And then also, as they’re building out use cases with these customers. They’re saying, hey, this, this new feature that we added, we just added pivotal Kubernetes service sounds like you guys could be using Kubernetes, let’s add the service on so it’s good for contract expansion. That being said, I don’t know that there are too many seed stage startups that want to start with the services model. But I think it’s an interesting way to think about monetizing down the line with open core, when we talked about the tweet earlier is a little bit of some of the backlash that people are having against folks that are commercializing open sources, basically saying part of your source code is closed part of its open heart. That is, I think, the way that I like to think about these businesses, and they’re not always built the same way. But sometimes the idea behind having part of an open source is a little bit of an insurance policy for the customer. Because they say, hey, well, if your company goes under, at least I’ve got this open source stuff, and I could try to run it myself, I can host it itself. Certainly you can try to build a robust community. And I think, again, I’m not sure that that’s so much about free labor, and maybe more about just kind of adoption, when projects get really popular among developers, and they bring them into work. And they say, Hey, why don’t why don’t we use NPM, it seems like that’s a much easier way to manage our JavaScript packages, right? All of a sudden, NPM is ubiquitous, the closed part of the open core model is important because it’s possible to run and build a project that’s so successful that you can’t afford to maintain it anymore. Sometimes people lose sight of that when they sort of get upset about an open source company commercializing that it actually does take money to do most of the things that are not writing code that these companies need to do like hosting, hiring, doing marketing and conferences and that kind of thing. So I do think there’s a pitfall for companies that open source too much and make it essentially impossible to monetize. So the question is kind of like, what can you open, you’re dealing with this community in open source, right, that has some very core values, I think it’s important what you close, sometimes people take the approach of let’s closed source, the really proprietary deep IP, sometimes people will say, let’s just close source, the part that’s strictly monetized, that maybe isn’t even the important part of the business to say the hosting platform or, or this kind of thing or like a dashboard, or kind of the platform where you interface right. So for example, a lot of Heroku is closed source, the build packs are open source. In theory, if you wanted to get off of a Perl, you could rebuild a lot of the stuff that they’ve closed source and use the build packs and get a lot of the benefit. Of course, it’s kind of a pain to do that. And it’s usually easier to just pay other people to do the hosting for you. And so I think that developers tend to not be super upset about that. They know that hosting costs money, they’re willing to pay money for hosting third model.
Erasmus Elsner 23:45
You mentioned the Saas model.
Lee Edwards 23:47
Yeah. So sorry, I kind of bled between open core and Saas, right. A great example of a SaaS model, I suppose would be something like segment or exit of New Relic. Yeah, these are interesting businesses where you know, and I could be wrong as possible. And they’ve open sourced some of that stuff. I think that a lot of the open source for things like New Relic is really like the client libraries. And the secret the secret sauce is very much closed, you more or less interact with that entirely on the platform. And there’s really no way to turn right if you’re not going to use New Relic comm you’re not using New Relic, right? There’s no again, there may be some API’s to pull your data out of the New Relic APM possibly, but that’s not what people are doing. They’re living inside a web browser. They’re living on the platform. So those can sometimes have kind of a walled garden approach. I feel like they work like they work really well for things like the BI tools and insights and analytics. There’s something good examples. I think, like the circle ci interface is really useful for people to get hub interfaces really useful. Do you think there’s something remains to be seen for companies say like weights and biases and folks like that, that are building I’m a huge, I’m a huge fan of weights and biases, and honestly kind of a huge fan of really any developer tool that’s out there, but they they have somewhat of like a walled garden approach. And you’re seeing a lot of folks that are doing kind of more command line driven, open source driven kind of alternatives to to weights and biases and we’ll see where that ends up. I think my bias as I think about like what I like as a developer, and when I try out new products and test them for what I call developer ergonomics, I tend to gravitate towards things that are sort of command line code first.
Erasmus Elsner 25:12
So maybe let’s segue to an idea that you’ve brought up in the past on raising the floor. And basically anything that provides the tooling to turn a 1x engineer into a 10x engineer, what used to be hard is now easy. And what used to be easy is now for free, I’m quoting you there. One example that you mentioned there was that in the past, people wanted some CMS system on the internet, they had to build their own CMS. Now they can use WordPress, if you want it to do credit card processing on the internet, you have to build that in yourself. Now you can do it with a couple of lines of code using Stripe, the same with AI, computer vision and TensorFlow. In your perspective, what’s the next frontier? There’s a lot of talk about the no-code and low code movement. Is this the new frontier?
Lee Edwards 25:55
I think of there being kind of two frontiers. If you sort of take the trend that I described as a given, the new code frontier is essentially what can you do with with no code whatsoever? That floor is rising all the time, right, I think we’re seeing YC companies that see multiple YC companies that are trying to make it so you can build mobile client apps that without any code. So you can think of it as like, Interface Builder that actually executes and runs the mobile app. Yeah, certainly, we’re seeing more and more things that are in kind of the web, the website builder space, and seeing companies that are trying to build you eyes, even for for building machine learning models. So I don’t know which ones of those will work. But I think there’s always this world where as problems become more and more solved problems, as we call them, you can sort of use really clever UI UX concepts to try to make them accessible for everyone. I do think there’s one thing that a lot of people forget about this floor, though, which is that more people can write code than you think I don’t like to dismiss to basically say, in order for something to be usable by everyone, it has to be no code, people who have journalism degrees, no HTML and CSS, they have to, they use it in their job, you’re seeing more and more of this. I mean, really, anyone is doing some amount of computer science to graduate look at the number of people that are writing macros in Excel. But these people aren’t really being told that they’re programming. They don’t really identify as like being a programmer. But their brain is working in a logical way, doing what programmers do. They’re identifying a problem. They’re breaking it down into its components. They’re applying their own knowledge, their domain knowledge. So I’m interested in that I do like no code. But I but I, again, find myself a little more gravitated towards, what can you do with just a little bit of code. And we have a company called daily that has this very simple API that allows two way real time video communication. So we’re on Skype now and it’s been performing really well so far. But as you know, it’s it’s it’s actually not easy to do a reliable video on that takes into account compression and latency and, and quality and all this stuff. That’s one example. I think Twilio and stripe are other examples where if you went to a boot camp, and you learned kind of basic programming in six weeks or something, and then someone showed you the Twilio the stripe or the daily API, you could just do it. And that’s kind of a crazy thing to move out to a video and web browser, like, say, five years ago. On the other end, like what’s the frontier at the top, I think there’s a lot of really interesting software technology that’s being built that’s enabling us to do things that we never could have done before. And this stuff is, you know, 10 years away from becoming no code or low code. It’s this point, very high code and requires, like, really brilliant computer scientists. One thing I look at, I mean, one thing we all look at a lot is, you know, stuff coming out of open AI, I think they’re doing some really interesting stuff. So I’ve been interested in simulation based training. So a lot of these machine learning models, like obviously, one of the huge problems in machine learning is access to large volumes of high quality data, and often well labeled data. So lots and lots of techniques about generating synthetic data about training things on simulation. So skydio, for example, right is doing a lot of 3d simulations as part of training their their AI algorithms. And so I’m interested in that I’m interested in a lot of the RL stuff that opening is doing their Dota AI and the way that they’re doing RL based on manipulation. One thing that we talked about internally, as I said, I’m focused purely on software. But I’m often talking software with my partners that are doing robotics, and they’re talking to me about software and AI. I don’t know that I can speak for the rest of the partners. But I personally believe that the only venture fundable robotics companies from here on forward will have deep AI on the founding team. And I think that’s because this sort of procedural way of instructing robots what to do, it’s not scalable anymore, and it doesn’t allow them to be adaptive in the environments. So when I was at iRobot, we worked on the PAC bot, right. And it’s an amazing product. It’s one of the highest selling military robot that’s out there. It does Explosive Ordnance Disposal, but it’s essentially a remote control car Explosive Ordnance Disposal EOD team, they got to be experts at operating this thing. I’m really interested in robots where you don’t have to be an expert. Certainly skydio is in that category. What’s going to enable that I think it’s going to be a lot of really sophisticated algorithms that are not procedural. They’re they’re learning model based. So yeah, so I’m always interested in stuff like that. The interesting thing about investing in this part of the world in this sort of category is I think it’s my job to kind of understand these technologies and figure out, you know, what they’re sort of capable of now and what they could be capable of in the future when founders are solving worthwhile problems using those technologies is when we’ll invest rather than sort of saying, Hey, I’m going to go out there and look for The next reinforcement learning startup, I’m kind of expecting and hoping that because reinforcement learning is so powerful, someone’s going to come up with a crazy idea that wouldn’t have been possible before reinforcement learning was was actually a feasible technique. And again, I don’t know what that idea is or found that company. But last time I talked about this, I had a really amazing RL founder reach out to me, so I’m hoping I can get more of that.
Erasmus Elsner 30:20
Yeah, We sure hope so. So let’s talk about your most recent public investment. And before we started recording this show, you were texting one of your founders, on whether or not he’s fine with announcing it. Maybe we have an answer by now.
Lee Edwards 30:37
He just texted back, go for it. Yeah, I had an investment earlier this year that I can’t talk about yet. But believe me when it does get announced, there’ll be screaming about it from the rooftops on Twitter. Yeah, so my most recent investment is in a company called Superconductive. So they’re essentially commercializing their open source project, which is called Great Expectations. So the founder of Gong just gave a keynote at data Council in New York. And the open source project has a very simple idea, which is basically you write tests for your code, right? We all think that code ought to be tested and documented, we think API’s ought to be tested and documented, we can argue about how deeply tested and how well documented but everybody believes that, hey, if I have this software, and I’m telling you to interface with it, I need to give you something a little more than the code data is more and more the interface these days, right? Certainly with graph qL, understanding the data models of the service that you’re interfacing with is very important. And even just internally within a company, if I’m, you’ve certainly if you’re in a microservices architecture, but even if you’re not even if you know you’re using a table that someone else created, and you want to know, like, what what is the temperature in Celsius or Fahrenheit, right? That’s a simple one. But there’s, we it, those of us with like, real world experience, we know that it gets can get way more complicated than that. And so great expectations is just a simple way. It’s almost like an R spec or a jasmine or something like that, to apply expectations to your data. So data at rest data in motion, you could you can imagine using it on a development machine, pretty much every place in your infrastructure. But I think that the biggest version of this idea would kind of the simple open source foundation that I think I think almost every company, almost every project could be using great expectations, it could be as popular as Jasmine is for JavaScript. Or I guess Jasmine is not even the most popular. I always talk about Jasmine, because it came at a pivotal. So. But I think the biggest idea here is that your data like throughout your organization, from product engineers, to business analysts, data science team, data engineering team, even eventually, like the finance team, ought to have kind of unified understandings of what is the state actually mean? What is this data lineage? Where did it come from? How has it changed over time, and a lot of that stuff probably could be inferred. I think that you could ask engineers to kind of spend time essentially like writing those unit tests, per se. But a nice thing about data that I don’t think is quite true of code yet, is that inference is very possible. So if I, if I have a stream of data coming in from a sensor, and then it’s piping into Kafka, and it’s training a machine learning model, I might notice like, hey, before it gets into the model, it looks like you have normalized this data between one and zero, or maybe you’ve done like one hot encoding or something like that, you could kind of automatically detect what’s happening to your data, and sort of make the burden of maintenance lower. And so what’s interesting about another thing that’s interesting about that is I think we all have documented our services and documented our data. One big problem is that that stuff becomes stale. So I look at things like swagger, and it does API documentation that’s actually based on your code. swagger does a not a perfect job, but it does a much better job of trying to keep your documentation actually up to date with your code. I think proponents of test driven development would tell you that’s exactly what tests do. So what if we could do that with data? What if we lower the barrier of maintaining the documentation of your data? So I think I think there’s a lot there in terms of
Erasmus Elsner 33:46
The founder of pachyderm, Joe Doliner (JD), on recently, which is doing version control for large datasets. And so I really hear you in terms of getting to grips with these large training data sets. Let’s wrap things up here. So Lee, thank you so much for being with us here today. Where can people find out more about you and find out what you’re up to? I mentioned your Twitter account, which is really the perfect combination of like private tweets, political tweets, startup and VC tweets. I’ve been following you for almost a year now. I almost feel like I know you as a person, which obviously is not the case.
Lee Edwards 34:20
Yeah, I mean, I think you might find even a little TMI about me on Twitter. I don’t know if anyone wants to know me that well, but that’s kind of been my philosophy is I’m a little bit. I mean, I’m just completely unfiltered on Twitter. So it is kind of like having a conversation with me. Yeah. So my Twitter is at terronkand I certainly take DMs there from founders all the time. You can also get ahold of me easily Lee at Root.vc, and I’ve got a small profile on Root Ventures. I also set up my personal homepage on lee.af recently, so there’s not much there. But I just wanted to get the domain because I thought it was awesome.
Erasmus Elsner 34:59
What does AF stand for actually in the TLD? Yeah, the top level domain, I mean.
Lee Edwards 35:01
It’s actually Afghanistan. Okay. Yeah, I hope your listeners know what AF means. But yeah, the TLD AF is actually Afghanistan, which is kind of interesting.
Erasmus Elsner 35:12
I’m surprised I haven’t seen more people doing .af top level domains. Lee, thank you so much for taking the time out of your busy schedule today and talking to me about these really interesting topics. I’m sure to follow your journey going forward with your investments. So thank you so much for this.
Lee Edwards 35:15
Awesome. Yeah. Thanks. Thanks for having me.
Erasmus Elsner 35:17
So this is it for today, guys. I hope you found it useful. Lee is such a smart and engineering focused venture capitalist, I think we can all learn a lot from him. And if you want to learn more about what I’m up to, you can always subscribe to my newsletter on sandhillroad.io or you can just subscribe to the channel here. Cheers, guys.