This is a great interview with Adam Kell. Adam is an investment partner at Comet Labs, which is a venture fund that invests in B2B robotics and AI startups. Their portfolio looks almost like a scifi movie. Because of this, Adam has an important perspective on where machine intelligence is headed. They’re helping to seed the next revolution of intelligent machines.
Adam’s background is in hardware, including designing capital equipment and processes for the semiconductor industry, including many semi-automated solar cell manufacturing pilot tools. He was named Forbes 30 under 30 in 2014.
Here are some things we talk about:
-Why is focusing on B2B robotics and AI advantageous to you and Comet Labs?
-How do you help your portfolio companies?
-What technologies have made robotics more accessible, easier to build?
-What technology still needs more developer before robotics truly takes off?
-What do you look for in an investment?
David Kruse: Today we are lucky enough to have Adam Kell with us. And Adam is an Investment Partner at the Comet Labs, which is a venture fund that invests in a B2B robotics and artificial intelligence startups. And Adam’s background is in hardware, including designing capital equipment and processes for the semiconductor industry, including the many semi-automated solar cell manufacturing tools. And he was named Forbs Under 30 in 2014. So at Comet Labs, which is a great name by the way, has invested in a number of robotics and startups. They are really helping to see the kind of next revolution of intelligent machines. So I asked Adam to be on the show to learn more about his background, and to hear more about his thoughts on robotics and AI now and for the future. So Adam, thanks for coming on the show today.
Adam Kell: Thanks Dave, good to be here.
David Kruse: Definitely and so let’s – I’m excited to talk more about what you guys have going on now, but let’s hear a little bit about your background before we jump in?
Adam Kell: Sure. So I studied Mechanical Engineering and Product Design as an Undergraduate in Upstate New York. I’m originally from the San Francisco bay area, going out to New York for college, then came back to the bay area to work for a solar company for a couple of years doing mostly capital equipment design. I went back to graduate school at Stanford to study Material Science Engineering and met my Co-Founder for a company that I then founded, which was consumer electronics startup and we designed and manufactured consumer electronics. We scaled that out to several products that were for sale in camping, backpacking, outdoor type markets. And then I did some work with an accelerator program called StartX in Palo Alto basically focused on educating Stanford’s best Founders and I ran the hardware program there, so I worked with about 50 hardware companies during my two years tenure there. I got to do some exciting things around helping companies find manufacturing partners and sort of learn about design firms and really figure out how to go to retail with the hardware profits and then after that I reconnected with someone who I met in college who was just starting up Comet Labs and sort of got going from there and joined the team in September 2015.
David Kruse: Interesting, that’s a great background. So I got a couple of questions. One is just a kind of random one. So how many start-ups do you think come out of Stanford every year, because you said you had 50 in just two years, just in hardware?
Adam Kell: Just in hardware, yeah. So StartX itself graduates about 100 to 150 companies per year now. They do three terms of three months each and then in terms of the acceleration rate at StartX is about, I think it’s less than 10%. So you can imagine getting to know 10 times that many companies in the Stanford ecosystem itself. They are other resources at Stanford University for those companies and obviously also outside of Stanford, but I would imagine on the orders of you know several thousand companies probably start.
David Kruse: Come out at Stanford. Wow! That’s crazy okay. I mean I knew – of course everyone knows they are amazing, but part of it is volume I guess, but…
Adam Kell: Right.
David Kruse: That’s interesting. Okay and I’m curious, what type of consumer products did you guys design and sell in your previous company?
Adam Kell: Yeah we did a device that generates electricity from waste heat sources. So you can take our device and fit it up next to your camping fire cooking stove and it would generate a steady USB charge for your mobile phone or for your rechargeable batteries.
David Kruse: Interesting. That’s a good idea. Is it still for sale?
Adam Kell: Yes, it’s available at stowerenergy.com.
David Kruse: Stowerenergy.com, well that’s a good domain, okay nice.
Adam Kell: Yeah, you can see a few of the products that we have for sale.
David Kruse: Interesting, okay. And so what in your past, in your background, why did you decide to – I mean you kind of, you have a very interesting entrepreneurial background when you decided to join Comet Labs and kind of focus on the helping nurture companies. I know what you did at Stanford was StartX, but what kind of prompted you to kind of get on that side of things.
Adam Kell: I think the main thing was this is really a good vehicle. The two most interesting things about the potential company that we invested now are (a) kind of the new technologies that are incorporated in it and this is sort of some of the most cutting edge technology that’s going into AI and Robotics. And then two, there is a lot of new business models that are enabled by the types of data streams that these companies are generating and so that’s pretty sync and really interesting ways of approaching go-to-market.
David Kruse: Well, that makes sense, okay. And this way I mean you get to kind of sit on top and then see what’s all going on from a higher level and see for patents and you know at the very cutting edge. So across different industries and technologies which is you are in a pretty nice position that way.
Adam Kell: Right.
David Kruse: So yeah, I mean lets – can you tell us a little bit about Comet Labs, how many companies you invested in and if you disclosed how large your fund is and yeah, we kind of talked about your focus I guess, but if there is anything more to add there?
Adam Kell: Sure. Yeah, so Comet Labs we focus on AI and robotics companies that are transforming traditional industries like agriculture, healthcare, logistics and retail. And we have done about 25 investments so far and that’s across both application based companies. So companies who are building a product for retail store like Target or for a distribution center like Amazon and then the other sort of the half of the portfolio companies focused on platform technologies. So more sort of fundamental building blocks than other technology companies can build solutions with; those are kind of like natural language processing, computer vision, universal robotic parts, things like that. Those are the types of companies that we invested in. Again about 25 in our current portfolio and the fund size is a $20 million fund. We’re investing from an evergreen fund right now, which basically means that we don’t have sort of a fixed amount, but about $20 million is what we are investing from right now.
David Kruse: That’s a pretty good amount. How did you guys raise that?
Adam Kell: So we are basically a spinout from an existing early stage we see called Legend Star and we were sort of focusing on geographic region and in the San Francisco area. We do investments outside of San Francisco, but sort of we are headquartered in the bay area and then doing these really focused investments like a very sort of, at least a very narrow subset of what the sort of convention Legend Star companies would look like and so we sort of just decided to rebrand and go out there and focus on these types of investments.
David Kruse: Interesting. Okay and what’s the typical stage that you invest in?
Adam Kell: Most of the investments that we have done so far at Seed and we have done a couple at Series A and then we have done one in Series B. But our sweet spot is definitely Seed and increasing as we raise more funds in the future, focusing a little bit more on Series A as well.
David Kruse: Yeah, that makes sense. Do you think you will always focus a lot on Seed even if you raise $100 million?
Adam Kell: That’s something that a lot of venture funds have. There is a lot of different ways that they approach that. Sometimes it’s hard to do Seed investments. We have a large fund, obviously for one reason because it’s hard to deploy very much capital. But also because it’s harder to support, like continue to support portfolio companies as they grow and not support others. Like if I invest in company A and company B and I keep investing in company A, but stop investing in company B, then there is a lot of signaling risk to other investors and so that might be another reason why larger firms choose not to do Seed around investments or just choose not to have that as one of the primary things. To me we are a Seed first investor and as we grow into the future that will be something that we have to think about, whether it’s you know either separating by creating a different fund which is managed by different people or you know whether that we sort of want the sweet spot to stay in the Seed, which would mean that we probably wouldn’t raise a $100 million fund.
David Kruse: Got you, okay. And when do you personally like to work with the companies, at what stage?
Adam Kell: I definitely like companies that are fewer than five people.
David Kruse: Okay.
Adam Kell: Being on more hands on with the team. I was approached to invest from a very product centric and technology centric view and I think that’s might favorite time of working with these types of companies. I think for other types of investments it, I don’t have any strange work, doing like for example consumer internet investments, but it seems to me like digging into the technology is a little bit more concrete than sort of you know being able to identify a company like SnapShort or Facebook in the Seed ground, which seems to me like a much more, I don’t know, a much harder process to try to identify those companies rather than just saying, okay you know, there are great founders working on this particular type of sensor and we believe that the sensor can be built and you know this team has a good shot of being that. The folks who built that senor that seems to me like it might end up easier than somebody who is doing consumer internet investments, but that’s just coming from someone who hasn’t been on that side of things before, but yes, technology and smaller teams are the two things that I look for in terms of the most exciting companies.
David Kruse: No, that makes sense, and how – it will be fun to hear some more about your portfolio companies that everyone should go and check them out, because it feels like you are in a Sc-Fi movie, which is good when you look through the list; flying cars, and robots climbing stairs and well not flying cars, but flying vehicles or flying, yeah drones and stuff. I’m sorry, not flying cars. We are not there yet, I guess. No, but you know can you give an example of how you worked with a company and like how you kind of provide advice or introductions and I know you have a lot of portfolio companies, so you know you probably don’t give everyone the same experience, but yeah.
Adam Kell: Yeah, I mean it definitely depends. You know we have done investments over a small part of a very large investment round. With those teams we are generally a lot less hands on, both because the teams are sort of more – usually typically a little bit more established and also because they have a lot of other investors who are trying to be providing value as well. But we have done investments where we have basally been the first money into a company and those companies sometimes even you know come and work out of our office space. We have a couple of those companies now and have sort of more direct, you know sort of more direct input, both in terms of strategy, giving feedback around pitch and things like that and then also just you know portfolio companies scraping our window and probably saying hey, could you introduce me to A, B and C; those are sort of the things. I mean more structurally the things that are were building, I had a sort of moment of realization when we basically – you know I mentioned we focused on traditional industries and how AI and robotics disrupt those industries. One of the – for one of these companies in particular that had raised you know quite a bit of money and I’d sort of contacted them a bunch of times and been like hey, you know what can we do that will be helpful and there are sort of ways like nothing really like, maybe we have those open hires that we are trying to make. Could you, you know refer to them to anybody or get them published on list, you know not really super hands on stuff. But then we, you know in the course of doing research around these traditional industries, specifically construction in this case, we basically hosted an event with a bunch of decision makers, about 60 people from large equipment manufacturers, general contractor companies, sub contractor companies, construction project owners, like basically all the stake holders in the industry and invited this start up and basically their feedback was like, Wow! This was like one of the most like focused events that I’ve ever been to. This has been super useful to us and kind of the takeaway for me there is, it wasn’t like they asked us to do that, because it’s like that’s not something that you I guess ask your investor to do, but it’s something that you know in our process of sort of becoming expert at this industry or at least bringing ourselves up to a high level of understanding to the industry; we were able to do some of the things that provided value to the portfolio of companies working at the construction space. So that would be like another…
David Kruse: Interesting. Yeah, I mean whenever a new technology comes through in the construction, that could apply in the construction industry you guys have a nice network of people you can make introduction to quickly, that’s smart.
Adam Kell: Yeah, absolutely. And we see a lot of you know cross industry vertical plays in terms of computer vision. Just a lot of fundamental tax can be applied to basically all the industries that you can imagine and that’s really why we think our focus works, even though you may look at it as on the service and say oh! They are thinking right out there, they are thinking about construction and manufacturing and logistics, but in reality a lot of these changes are being powered by the same core technology and those are the things that we want to become expert in.
David Kruse: Yeah, that’s a good point and so can you expand on that a little bit. I mean you mentioned the computer vision. Can you give an example? I mean if this is like part of your core investment thesis, you know to give away the store, but yeah, can you?
Adam Kell: Yeah, we have a couple of different portfolio companies who focus on very different applications, but are just two really good examples of how sort of important these technologies are going to be in the future. One of them focuses on computer vision. They do facial recognition for financial security application. So like if you go to an ATM and you swipe your card, it basically knows who you are and they can use the canvas to basically make sure that that person is the one who is using the device and the way that that works is basically it has a learning systems or basically it takes in data every time you either use an ATM or you know pay for item and there is a camera system there. It takes a lot of pictures of you, so even if your appearance changes, your hair grows longer, your grow a beard, it can recognize you anyway and sort of using this deep learning network that they have developed to be really good at facial recognition. So that’s one application on the financial securities side. And then another application is a company that we recently invested in who does this special device that can basically biopsy very thin slices of tissue, of human tissue and imagine it at the same time. So you can take like a tumor for example and create a very like paper thin slices and actually biopsy and image the tissue at the same time and with that they use computer vision algorithms to be able to figure out why one person’s cancer is different from another person’s cancer on a cellular level and then from that we can certainly make, you know make some more progress on understanding cancer for example.
David Kruse: Well, that’s pretty brilliant. So those start-ups every talk to each other about their problems that they are having?
Adam Kell: I think today not as much. But I think for one we are interested in that being one of the value adds that we provide in future, it’s really creating more of a studio model for people who are using similar technologies across different applications, to be able to have kind of this venue to be able to talk about technology and obviously you know it’s helpful when a lot of people are trying to recruit computer vision engineers to be able to recruit them ourselves by having you know kind of this studio model where people work on the things and people can you know – engineers can come and work on problems and use that as a recruiting tool is another things that we are thinking about.
David Kruse: Interesting. All right, so you mentioned I guess two or three portfolio companies that are there and I really don’t want to pick favorites, but are there like a couple of other portfolio companies that you really – you’ve kind of seen blossom or you’ve been with for a while or your excited about?
Adam Kell: Yeah, I think well there is a rule in venture capital, which I think is also there’s a rule in parenting as well, which is you are not allowed to say which one’s your favorite or the correlated that is my favorite portfolio company, the one that’s fund raising with the most urgency. So I’m going to break that, I’m going to break that second rule and tell you about a company that’s absolutely that’s not fund raising right now, but I think illustrates kind of an interesting application. Kind of I’ll tell you about the solution and I tell you why I think it’s interesting. Basically what they do is they put a sensor in a retail environment that has a cameral microphone and they basically use computer vision to be able to figure out how many people are walking past the store outside on the sidewalk versus how many people come in, and they also look at interesting statistics about how many people – like how long people stay in the store, like dwell time. They use their microphone to roughly correlate like to basically back up that data in terms of how many people are in the store in terms of ambient noise. And then what the company does is they basically take this and connect it with the data at the registers; like how much you know on average did each customer buy that day? What was their basket size? And what percent of them converted versus you know maybe 10% of the people in the store bought something one day, but 30% bought something from the store the other day and what it does is instead of giving a dashboard to the store manager, it actually just manages the schedules of the sales people themselves and what it does is kind of randomizes or shuffles themselves. It correlates which sales teams are the best at being together.
David Kruse: Wow!
Adam Kell: So say you and I are on one sales team and me and Jim are on another sales team. Maybe Jim and I don’t work very well together, but you and I work really well together. Like we sort of consistently when on the floor together have you know higher output and that’s something that the system learns over time and then will preferentially schedule us together. Then there are other things, you know there are other things that they conclude in terms of weather and day of the week and like how many customers we expect to have and how many sales people should we have, because those are pretty expensive assets to be either under or over utilizing. For example like to – if you and I are the only ones on the floor and there’s 15 customers, obviously like I don’t know if you’ve ever been to Foot Locker and waited like 20 minutes for somebody like to ask you like what you need or what size or whatever, but that’s not like a great way to drive sales. So really focusing on balancing demand and supply and then also optimizing sales teams are the two things that they focus on.
David Kruse: Interesting. Yeah, I mean when you first started talking about it, I was like, Oh! There’s kind of companies that do that, but then yeah that’s another whole level they are bringing it to where they are trying to optimize the sales staff and the staff. Well, that’s cool and I can see it…
Adam Kell: Yeah, and I think – I mean the reason why featuring any of it is like it’s a very subtle trend between like saying, Oh! We know how many people are walking past the store versus how many are coming in and therefore we can like change the display in the front or the signage or whatever versus like kind of going a few layers deeper and saying like, no lets actually like change the, you know change like which sales people are on the floor at which times and sort of optimize it that way and it becomes a much more passive thing that the store room or the manager doesn’t even have to think about. It’s just a system that’s running in the background and I think that’s illustrative of a lot of these artificial intelligence systems are going to be these agents that sort of run in the background and take away a burden as opposed to just giving us more information.
David Kruse: So do they somehow connect with the scheduling system? Is that – or do they have their own scheduling system for that?
Adam Kell: Yeah, they actually just do their own; they just manage the scheduling themselves.
David Kruse: Interesting. And can they track people going through the stores? I don’t know if you said that or not. I know that some companies are working on that, which is a little questionable, but yes.
Adam Kell: Sure. Right now it’s just focused on the inverse of the doubt and like how long people stay in the store. I mean you can imagine a future where you know everything about every customer that walks around the floor and is looking at – you know is looking at something from a – on which shelf in the same way that Amazon has all that data today about how you shop on Amazon and they know what’s your search. What you kind of scroll through and click on? What you read in the description and what you expand in the comment sections or read more comments. They know what you add to your cart; they obviously know what you check out of your cart and they kind of use all these things to go into that model that determines customers who view this also viewed or recommended next, next purchases. Netflix does the same thing and a lot of Brick and Mortar retailers are trying to figure out exactly what – you know how deep in the stock do they want to go and to your point which is, is it creepy? Well, its creepy that its being done online right now and customers don’t seem to think its creepy, but it does seem creepy in a Brick and Mortar retailer, because there’s a bunch of cameras on you and sensors that trigger out what you pick up and read the back of the box and stuff like that, but yeah I think sort of consumer appetite role and privacy role will sort of be one of the big discriminates of like how much of this technology will be adopted and how much the equilibrium will sort of switch to being e-commerce driven versus Brick and Mortar driven.
David Kruse: Yeah. I mean I think people will get use to them. I mean it will be pretty awesome to walk in the store and it’s kind of like Dave’s store, right. Everything that’s – kind of like you said with Amazon. It goes Dave’s store, you walk into target and everything is – not everything, but a lot of the displays will be customized for me and the suggestions and how people ever talked, yeah so that will be cool, someday, someday.
Adam Kell: Yeah.
David Kruse: All right, well let’s see. I was curious to get your thoughts a little more on robotics, because you know there is always the promise of robotics will take off – and they have in like manufacturing and certain areas. But you are looking a little more in the non-traditional areas like healthcare and yeah. So how– what type of do you think B2B robotics will be successful in the near future. You know like what do you think people should be working on right now. You know they are not going to necessarily go replace a human right away, but maybe they could. But yeah, do you have any thoughts on that?
Adam Kell: Yeah, so I think one of the big things about getting robots off of a manufacturing line, which is like a very, very structured environment is solving the perception problems. So actually being able to tell like if – you know if you imagine a car frame moving down a conveyer belt and a bunch of robotic arms welding that, that’s a very like structured environment and the robot arms are programmed to basically do the same motion over and over again. But moving to a more sort of a structured environment like say picking apples for example, there’s a lot of problems in trying to figure out like where the apples – you know the apples aren’t coming down on a conveyer belt. They are on a tree in different connotations and sizes and shapes and colors and they might be included by leaves that are covering them and being able to figure out – for robots to figure out how to actually get a manipulator up there, get the leaf out of the way and figure out if I should pick the apple or if I shouldn’t and then how to get the apple off the tree without damaging it, it’s either a sufficiently hardened problem, but have only been able to be solved very recently and I would say a lot of that comes from the perspective of being – having the perception problem solved, which is like being able to actually understand the environment as it is today. Like looking at a tray and saying, okay I know where the apples are on that tree. That’s been one of the problems that I would tell you is one of the biggest items that’s in fault for a lot of these. But why wouldn’t they solve this? It’s been solved through a threshold that makes – you know we can make useful solutions today and we will be continually making more useful solutions in the future and mostly based on perception.
David Kruse: And why do you think that’s been solved – well, at least some of the problems have been solved now in the last, let’s say a couple of years versus 10 years. Why is now special?
Adam Kell: I think it sort of like – this is like a lot of the main sort of argument. So anybody who gets excited about robots says too many things, which are like robotic components are getting cheaper, the cost of computing is also getting cheaper and moving to the clouds that you don’t actually have your own infrastructure and then kind of this ubiquity of cameras. Basically being out there and cameras being super cheap and cameras being super good; basically all these sort of complement of things, the sensors, the computing, like everything is getting cheaper and better. So like fundamentally cameras today, thanks to Smartphones are like so fantastically better. If you like look at the old, like pictures from like an old – like the first iPhone that came out for example, like it’s crazy how much – like Apple in itself has 800 in queues just working on the camera module for the phone. It’s like the amount of improvement like how fast the technology is being pushed forward is pretty astounding and that’s basically how we can, literally like there is some more where basically one way to think of it is the cameras does not take in more data. Like this is a lot more things that are in the data, like a lot more pixels and from that you can do a lot more things, which is like you know find the edges and decide whether this is an apple or a leaf and those are something that you know are just starting to become more possible today as compared to even a few years ago.
David Kruse: Yeah, I think I even heard that somebody is working on, one of the cell companies like a depth sensing camera or a 3D camera which would be – that would be pretty sweet. Of course there’s solutions out there, but once it’s on your phone that would be pretty amazing.
Adam Kell: Yeah, absolutely.
David Kruse: All right. Well, we’re almost out of time here. I did – I was curious, I guess how do you find your investments and what type of invest companies are you looking for, which we kind of talked about and we know the area, but at what stage, you know you’re at Seed. So do you want somebody with a good idea, do you want like proof of concept, you know at what point we’re like, ‘okay, we’re ready, Comet Lab is ready to put some money in.’
Adam Kell: Sure. To answer your first question, what do we find in most of the deals, just deep sort of networks. We focus really you know somewhat narrowly on robotics and artificial intelligence and because of that we don’t have to look at the consumer enterprise software companies, consumer companies, consumer electronics companies, like there are a lot of things we just don’t even have to look at, but these [inaudible] small where we can get a lot of the, you know a lot of our – the investment team can basically just hit the streets and be talking to the people at these universities or at these organizations we interface with the companies who are building the new technology, to be able to find the deals that we’re looking for. And then your second question is like what does the company need? What does a successful company that picks a smart class look like? There’s basically an inverse relationship between team experience and like the amount of traction that they have to have. But if it’s the guy or girl who literally wrote the book on some technology, we’re probably going to be investing in that company no matter what they are, even if they don’t even have a power point slide, yeah. But if you know it’s the first time Founder building you know a solution with off the shelf components and also there’s some problem in the industry, that thing is something that we – you know the bar is a little bit higher in terms of things that need to be demonstrated. When you think of that, that’s partially because of the risk involved, but it’s also partially just because we’re worried about the defensibility if somebody can sort of just go out and solve this problem pretty easily, then it’s pretty likely that other people can do that as well and so that’s something that we definitely take into consideration when we make investments.
David Kruse: Makes sense. And do you typically invest in the bay area or do you have companies outside?
Adam Kell: We have companies outside of the bay area. I have done a lot of investments in Seattle, San Diego, Los Angeles, Pittsburg and Boston.
David Kruse: Wow! Okay. Interesting and great. All right, well, I’ve got – I could talk for another hour with you, but I think that most of all does it for us and I definitely really appreciate your time Adam and your thoughts and the – yeah, well what you guys are doing, it must be – at least a lot of days must be fun waking up. I’m sure not every day is blissful, but it’s cool what you guys are working on.
Adam Kell: Yeah, I know it’s a privilege to have the job and be able to work with such smart people.
David Kruse: Definitely, definitely. So yeah, I definitely appreciate you coming on this show and telling your story and as always, I appreciate everyone listening to another episode of Flyover Labs and yeah, we’ll see you next time. Bye everyone.
Adam Kell: Thanks Dave.