Cloud BIology: Automation and the Future of Biology

Welcome to the Axial podcast. Axial is an early-stage investment firm based in San Francisco. We partner with great founders and inventors investing in early-stage life science companies, often when they are no more than an idea. Axial is fanatical about helping the rare inventor who is compelled to build their own enduring business. Hey DJ, how are you?

Good, yourself?

Doing pretty well, kind of enjoying the weather in Berkeley, right? I don't know. Seriously, Thursdays are my best day. I go to the Rose Garden over in Berkeley. But if you make me a moderator, you're new, so let me— if you click on my profile, you can say moderator.

I gotcha.

Is this room— yeah, people will show up. Okay, cool. Well, great to have you here. Really excited to just like have a conversation with you and just tell your story more broadly. And I think ECL has some really incredible infrastructure. So hopefully, you know, inspire some scientists and entrepreneurs to reach out and plug into ECL. But everyone will really know. I'm really excited to have DJ Kleinbaum, the co-founder. And is it CEO or co-CEO still?

Co-CEO. Co-CEO.

Very important. You and Brian are co-CEOs of ECL. Emerald Cloud Lab, uh, and it's, it's, uh, I'll let DJ talk about it, uh, but, uh, yeah, maybe we can— you can introduce yourself and we can go from there.

Yeah, absolutely. So I'm, as, uh, as Josh already said, I'm DJ Kleinbaum. I'm one of the co-founders of Emerald Cloud Lab, and what Emerald is is a remote control life sciences laboratory. So we've effectively built a software application that scientists anywhere in the world can use to design experiments as if they're standing in front of the instruments themselves. Once those experiments are designed and submitted to our system, they get run in an ECL facility, facility exactly according to the specifications that we've been given, and then the data is handed back to the scientist through the same software application, which also includes a whole suite of tools for analysis, visualization, plotting, basically figuring out what happened in your experiment and deciding what you want to do next. This has been a relatively long and circuitous journey for us. And Josh, I don't know how much you want to get into that, but we've been at this for more than 11 years now.

Yeah, I think good things take time. And I think Emerald has an incredible story. And anybody in the audience, if you have questions, this is more of a conversation. So I see some familiar faces. So always happy to have questions and a different perspective. But maybe we can start from day one. You know, how did you and Brian, the other co-founder, meet? And from my understanding, you, you're both in Philadelphia.

Yeah, so, so our story— I'm a horrible person to talk to about how to, how to like meet or find a co-founder, because the only experience I have with it is when you're 9 years old, have your family move to a house that's 2 doors away from someone who's going to grow up to to be one of the most brilliant scientists of a generation. So that's how I did it. I'm not sure that's particularly actionable for others, but Brian and I have known each other since we were 7. So since probably the first day of first grade, and then a couple years later, my family moved to a house on Emerald Drive, which is where the name of the company comes from. And, you know, from third grade on, we were inseparable.

That's awesome. And so, uh, Brian went to Scripps for chemistry and you went to Stanford. You both went to Carnegie Mellon together, right?

At the time, yeah, probably from the time we were 14 or 15, we were really interested in this intersection of computer science and the life sciences, which sounds obvious now, but I promise at the time, you know, made us very weird. I had someone who actually, when I told them that was what I wanted to go study in college, they asked me, like, why don't you study physics and French literature? They have as much to do with one another as computer science and biology. So the last, whatever, 21 years, I think, has proven us right. But at the time, Carnegie Mellon was one of only a few schools in the country that even had a program that allowed you to reasonably, like, study both of those things. Yep.

Also, one side note, you two both went to different high schools though. Was that to kind of not compete? You know, so both of you could be valedictorian, or was it just kind of happenstance?

No, it was, uh, it was just happenstance. In fact, I used to joke, so we went to the same elementary and middle schools, different high schools, the same college, different graduate schools, and then started a company together. So I used to worry that I was only tolerable for like 4 years at a time. Thankfully, that seems to not be the case because I, yeah, we're on year 11 now and, you know, some iteration of the Emeralds. So It's, yeah, so thankfully that wasn't the case, but we ended up going to, yeah, we did go to different high schools.

Yeah, and that's always really interesting. Sometimes I have friends who, you know, they're very, you know, they're trying to be number one at their high school, so they have another friend of theirs, they don't wanna compete. Seriously, I grew up in Orange County, and there's a high school called Troy High School, and it's like, I didn't go to Troy High School 'cause I wanted to be valedictorian, so I went to some other high school. Troy High School had so many smart kids, and I was like, let me go to a school Not so smart kids. And so, okay, Brian's at Scripps, you're at Stanford. What was the premise of, you know, Emerald Cloud Labs was Emerald Therapeutics. And so maybe you could tell that story of being in grad school and one, wanting to start a company and then two, realizing, holy smokes, we might have to build infrastructure for ourselves and other people.

Yeah, so I mean, I think we knew we wanted to start a company way back, like before we even— I think we both went to Carnegie Mellon knowing we wanted to come out of this and start a company. Brian's dad was a serial entrepreneur and a venture capitalist, and his brother was very important to the early days of PayPal and worked very closely with Max Levchin. So, not only do we have the startup bug from from hearing those stories and those experiences from the time we were young. But I also always kind of suspected that I was way too insubordinate to do anything but like start my own company or start my own business or anything because I just would not work in a sort of like larger infrastructure. And so, you know, we got to the end of our time in undergrad and had ideas for like a company that we would, you know, Many ideas for a company we would start, in retrospect, all of which were bad, but we started talking to other entrepreneurs and investors and everyone told us, "No one will fund or let you run a biotech company unless you have 3 letters after your name." So you hear that several dozen times and eventually you're like, "All right, screw it. We'll go to graduate school." So I don't have a lot of particularly positive things to say about my graduate school experience except that it gave us the better part of 5 years to vet dozens of different ideas about what this company was going to do, and not just about what we were going to do, but how we were going to do it. Because I think we both got to these top research institutions, and given our engineering and computer science background, we're just appalled at the state of research. And we would have these calls where he'd call me or I'd call him, I don't think the emperor's wearing any clothes. And it was very validating to have someone on the other end of the line and be like, you know what, you're right. It is ridiculous that we're using this scientific instrumentation that costs hundreds of thousands of dollars and generates this incredibly precise data. And the way that we save that is to print it out and glue stick it into a paper lab notebook. Like, that's crazy. And so by the time we finished graduate school and started Emerald Therapeutics, we had just as many ideas about how the company would accomplish its goals as what those goals were. And that ended up being very valuable because the summer of 2010 was not a great time to be raising money in biotech, especially if you were a couple of nobodies right out of graduate school, which we very much were. I still consider it to be a minor miracle we were able to raise any money, but this was not a gigantic Series A. This was not a flagship Series A. By any means. And so, we didn't have the funds to run this like a traditional 20th century biotech startup. And that was good because we had no intention to. So we had, we started building tools for ourselves just to give our small team the leverage that we needed to get all the necessary work done for this therapeutics effort. And then that just grew over time to this much larger and more holistic system.

That's really awesome. Yeah, I totally feel you on grad school. I mean, I love science, but doing biology is so hard and I have so, you know, honestly when I read science papers, I feel like Salieri or something. It's like, oh my gosh, you know what I mean? Like how did they do this? And that's why honestly grad school didn't work out for me because like bench work is way too hard. That's why, you know, Emerald and Transcriptic back in the day, you know, are so attractive cuz like if you can kind of decouple creativity from the actual, like, hand, hand, like, labor, uh, you might have more different people, uh, designing experiments.

Uh, right.

So I honestly, I might go back to grad school in 10 years. Seriously, again, I need to— I actually need a PhD. So help me get a PhD, JJ.

Listen, if we can make, you know, if we can, like, empower people to, one of the great tragedies, I think, of these sort of like graduate school, like academic complexes, people leave graduate school or leave a postdoc thinking the value they bring to the world is what they do with their hands, not what they do with their head. And that makes sense when you think about how your time is distributed as a bench scientist that, you spend a week planning out a set of experiments that you know is gonna take you 6 months to get to the end of. And anything we can do to sort of shrink the time from idea to result is, you know, doesn't just like make graduate school more palatable. It makes, it's better for all of us because it makes science go faster. It allows us to explore more and like more ambitious ideas because the labor involved is just, it's less onerous. I mean, I had this very, I think transformative experience when I was an undergrad, I took a graduate class in bioorganic chemistry. And one of the last things we had to do was everyone had to put together a research proposal. And all the undergrads who had, we all had done research in the lab, but we weren't grizzled veterans. All our ideas for these research proposals were things that were incredibly ambitious and would easily be if you made it work, it would be the fastest, like, uh, from like the paper to like a Nobel Prize that you had like seen. And then all the graduate students had proposals that were like, I'm going to take this thing that I understand really well. I'm going to take a tiny step forward. And I think the difference between those two is that the undergrads did not, we did not fully comprehend the amount of like labor required to like do these very ambitious projects. And the graduate students felt that very viscerally. So How could you do anything but try to move things forward incrementally?

I totally agree. I think we'll get there. And I think eSales is in the lead there. So maybe we can kind of move towards kind of the platform you have and so, and keep it on the timeline. So you have a therapeutics company.

Oh yeah.

And so one thing, I love hype, I love like, bragging about things. I love hyping people up. That's my job. And so ECL, it's incredible what you guys built. You guys have like hundreds of devices and then you have like thousands of different functions, you know, being fully integrated from, you know, taking that data from instrument, doing analysis, and then delivering, you know, some sort of insight towards it. And so, and in a combination of being fully integrated, you know, your labs can just, they're on every, they're on every day, 24/7. And it's just, that's another, you know, I live in Berkeley and sometimes I walk by the labs and I feel a little depressed. I'm like, man, it's like midnight right now and I'm seeing somebody like doing a PCR. Whoa, that stinks, man. I've been there. That's not a fun life. So if you can kind of, ECL is not only helping science, but it's also reducing suffering. And so, you know, what was the kind of like from day one, Or what was that moment where you decided Emerald Therapeutics, we're going to go build a lab too? When was that kind of switch?

Yeah, it was, so it was very organic. So, you know, at, as I mentioned, like at Emerald Therapeutics, we, the two, we did two things that were very different from the start. One was we invested a lot of the small amount of money we raised into capital equipment because our goal from the beginning was every experiment should be push a button, walk away. We couldn't afford to hire a technician and like camp one in front of each instrument that we had to keep running. And so we needed a way that our scientists would be able to just run many things at once. They would be able to be more leveraged than like your average bench scientist. So we invested in automation and not just for like high throughput stuff, but just for everything. Some of it's really like simple, just like buying auto samplers and fraction collectors for LC systems so that you don't have to wait for your fraction to come out, which I spent countless hours doing as a graduate student. And so that was, so the hardware piece was like one. And the other piece was we taught all of our scientists how to write software and not like low-level stuff. We weren't teaching people how to do cash validation and memory management or anything like that, just enough so that they could build tools to make themselves more efficient. So you could build a function for like doing, you know, doing plot analysis so you didn't have to like spend time like picking your peaks every time you had like run an LC. And that started as like very simple, just like build yourself tools so you can be more efficient. And over the course of 3 or 4 years, that slowly grew to encompass everything that happened in the lab, and the last step of which was the one that sort of just started thinking about the protocols that were run in the lab as just another node in this sort of data graph that we were— data graph or knowledge graph that we were building so that our scientists could sit at their computers and, you know, specify, you know, an experiment they wanted to run. The software would generate a machine-readable file that could get loaded onto the instrument. And then the only thing you had to do in the lab was the logistics, like, okay, I've got to move like this carboy of buffer to this instrument. I've got to put this plasticware on the deck or move a sample from point A to point B. And that was the only time you had to be in the lab was just sort of like shuttling things around. And at that point, just the productivity of our scientists just skyrocketed. And I actually realized this because I was, uh, going through our budget and our burn for a board report. And I was like, how are we, how did we spend so much money in reagents over the last month and a half? It was actually kind of like a terrifying moment. But then I realized it was because our scientists were running so many more experiments once these, once that protocol feature had gone live. And so at some point, around that time, we went to our board and we kind of said, look, we built this internal tool. There's more value in this thing than we can capture just by using it ourselves. This kind of needs to be its own thing. You know, we built this cloud lab internally. I think, you know, uh, Brian, my co-founder, and Stephen Wolfram started using the term cloud lab, I want to say, like, years before this. This was like circa like 2011. They started talking about what we were doing as a cloud lab. It was just a cloud lab that was internal to Emerald Therapeutics. And so, you know, after many conversations with our board and way too many conversations with lawyers, we eventually decided that these two things needed to be on their own trajectory. So we actually spun the companies off from one another. So Emerald Cloud Lab and Emerald Therapeutics exist as two completely independent corporate entities.

That's really interesting. As a founder, like, how did that that process go where, you know, Emerald Therapeutics, you're trying to solve one problem, make medicines for patients. And then it's then, oh, we have to solve our own problem. And you see that a lot in tech. I think Slack was like that. A bunch of tech companies are like that. Biotech, you don't really see that actually. You don't see like a drug company and say, oh man, we're going to spin out something. Maybe WuXi, maybe WuXi is like that actually.

Uh, but I mean, Millennium, Millennium, it's not, it's not exactly the Millennium story, but it's like if you kind of squint and tilt your head, it's kind of the Millennium story. Uh, but you're right, it's much more common in the, in the tech world than in the biotech world.

Actually, I think WuXi is like that. I, I think, yeah, I think if ECL essentially WuXi, you know, ECL, the big opportunity, I think I told you this, is like just to build WuXi in America. And, uh, seriously, and, uh, for, for national security and a bunch of other reasons. And so I think that's your tailwind or something. But, you know, so seriously, you can solve the labor arbitrage problem because you're tech-enabled.

Right. Right.

And so in terms of like, you know, you kind of started another company. How did you like balance your time? How did you reorientate your team? How did you like, you know, you're solving a different problem now with ECL. Right. Solving the problem of, you know, increasing reproducibility, lowering barriers, kind of what was the kind of just the organizational logistics and kind of that reorientation, you know, and how did you do that?

Yeah, so I'll address the team question first because that was actually, that one was really easy because it was a huge relief to our team when we, when we divided the company because prior to that I think everyone felt this great amount of like anxiety and pressure about like, what should I be like focused on? The first derivative of, you know, basically building the platform, building the ECL to make research go faster tomorrow, or should I be trying to push like research forward today? And so once we like drew that line very clearly of like everyone on this side is like building, is like on the tool side, and everyone on this side is like on the research side, and is not focused on building the tools, just focused on pushing the research forward. That was actually very clarifying for people and very motivating because then they could just have one point of focus to worry about and didn't have to be in the situation where no matter what they were doing, they felt like they maybe should be doing the other thing.

So how did you divide the people? Was it by straws or something? Like, hey, everybody picks a straw?

And it was actually, it was really clear, like there were just some people on the team who were much more focused on, much more focused on and motivated by the therapeutic. And they very clearly like went with that company. And then everyone else, like the vast majority of the workforce, was much more motivated by, you know, the tool and the platform and the infrastructure. And so it was a pretty clear division. There were a couple of people who I think could have gone either way, and we basically just gave them a choice at that time. We were like, look, you kind of need to pick a side of the fence to sit on. Where do you want to be? And that made it, you know, they made their choice and we never had anyone. I've actually talked to a couple of those people recently. No one regrets the choice they made. So I think that that ended up being all right. And in terms of Brian and my time, the reason why I think we've been able to make it work is just, you know, Emerald Therapeutics was very much still in the early days of research and development. And so, we have an incredible— the woman who runs research for Emerald Therapeutics is brilliant. She has a— the rate at which her brain spins is significantly faster than mine. I always say that Catherine has one of the highest RPM brains I've ever seen.. And so she doesn't actually need that much from us on a day-to-day basis. Like the Emerald Therapeutics, for, for me anyway, is like playing correspondence chess. Like I need to make several decisions like a month or like a few decisions a quarter. And those decisions I have to think about a lot and really focus on, but there's a small number of them. Whereas Emerald Cloud Lab is, you know, 1,000 decisions a day. So it's like playing speed chess. So because the timescales are so different, it's been palatable to, to manage both.

Really cool. And then, you know, when you start— when you kind of then sort of built— get ECL off the ground, how did you define the problem set? I think a lot of founders, you know, when they get a company started or early days, you know, just trying to define like, what are we going to really be the best at? And from my understanding, it's reproducibility, equipment access, reduce onboarding times, standardized data and a few other things. But how did you go about— were those problems obvious to you, or did you have to iterate across, you know, several months or something?

No, I think all those problems were obvious. I would say those problems are more the— those, like, those solutions are more the output of the, the central, like, problem that we are trying to solve, which sounds much more pedantic, but I think it's very important, which is how do you create an instruction set for the lab such that there's nothing that you— nothing is off limits. Like, there's nothing that you could do if you were standing in the lab yourself that our instruction set does not account for. So, you know, like, this is one of the reasons why it— we've had to be, you know, we've had to sort of like tear down and rebuild our system many times because eventually, you know, you get into the devils in the details. You're like, oh, like our, you know, we are our cut, you know, when it was the early days, it was just us, but now it's our customers need to be able to do like this thing. We don't have like a, an instruction for doing that. So you can almost kind of think of it like an early, like a microprocessor that we needed a, an instruction set. You know, you don't just build like a function for addition and a function for subtraction and what have you. You build an instruction set that you can compile those, you know, higher order functionalities from. And that's what we were trying to do is what is sort of like the instructions that you need to drive a laboratory remotely. And then all the other things that you mentioned, the reproducibility and the speed to start time and reducing like the overhead capital barriers and all of these things are all things that sort of fall out once you have that. But we didn't say we need to solve the reproducibility problem. How are we going to do that? We said, We need to have an instruction set so that you can specify exactly what you want the lab to do remotely from the jump. And then what are all the benefits you get from having that?

Really cool. Really interesting. And so from that basis of like a framing what you're working on, how long did it take for you guys to get that first version of the product out? And during that process, Like, who are you building it for? Maybe a wide set of people, or did you have someone in mind, like, we're going to build this for like an individual scientist or an academic scientist, or for at least that first product?

Yeah, so I mean, the first product, when the first product like came out, it's sort of a, you know, we had like an internal, this internal tool and We've constantly been, ever since the split, we've been iterating to make the UI easier to use and to make it more intuitive and simpler for people to use and to drive. So where V1 really starts is kind of a hard line to draw. But I think from the beginning, we set out to solve a problem for ourselves. So early on, we were really thinking about we were really thinking about startups. And we envisioned that startups and maybe academics would always be our first market. And then eventually, once people had used this as maybe as graduate students or in like a startup company they had done and then had gone to a big company, eventually through like a grassroots effort, we would get into big pharma and sort of larger enterprise customers. And it hasn't exactly been the opposite, but it's almost been the opposite. The amount of interest and hunger for this, at least among some segments of the sort of larger pharma and biotech community, have— they haven't, you know, outpaced the startups, but they've at least kept pace with them, which I never would have expected when we originally, you know, sort of designed this and thought about what our go-to-market would look like.

Interesting. And why do you think is that? Do you think it's because biopharma companies just have more data, more samples to process? Or is it just a, like, maybe is it a resource issue? Or is it a, is it kind of, sometimes startups need to do their own experiments just to raise venture capital and not seem like they're giving IP away.

Right.

And yeah, do you have any—

Yeah, I mean, I think it's different. One of the interesting things is that even though the product is the same for all of our customers, everyone sort of like sees or focuses on a different point of the value chain. So for example, for some of the earliest big pharmas we worked with, some of them were really focused on the fact that they described their work with their outsourcing partners as an existential crisis, that there was so much science that was not happening within their walls anymore, and they were effectively, they even as scientists were effectively just like PMing scientists like thousands of miles away saying like, do the thing that I would do if I was in the lab, but not really confident they could really understand the science well enough to do that. And so they saw Emerald as a way to take back the control of the science without having to lose all the benefits of, you know, elasticity and like no overhead capital costs and scale and all these things that they had with their outsourcing, outsourcing customers. For others, it was companies that really wanted to get into the data science and ML and AI worlds, but, you know, had tried and had just realized like they spend 99% of their time munging data to try to put it into a form that those algorithms can actually work on. And even then, like, it's hard to get anything actionable out of it because it's so hard to compare like apples to apples when you're just, you know, mashing datasets together. And so for them, the fact that all the data that gets generated in the ECL comes out in this pristine form in a very, like, precise structured ontology that can instantly get fed into any of these different algorithms. And, you know, not to go too far afield here, but we even have some groups in the computational biology and computer science world that are now trying to build, like, closed loops where they get results back from Emerald, and then they have an algorithm that does designs the next set of experiments and then just feeds them right back into the system. So there is some very cool stuff starting to happen there. And then for startups, I think that the— so for the big pharmas, it's really— there's a wide diversity of why they came to us early. And then for the startups, I think early on, our whether or not we resonated with companies was just really directly tied to whether or not we had the instruments that they needed. I mean, that, you know, there is this coverage piece which we're almost at the end of. We're not quite there yet, but we're pretty close. Is huge that, you know, if someone says, hey, I want to be able to— I need to be able to do these 7 different experiments or basically have access to these 7 different instruments. And we were like, cool, we have 4 of those. Like, all right, great. We'll come back when you have the other 3 because until I can have the complete set, I still need to rent and fit out lab space and buy CapEx and all of these things that just make the whole proposition interesting but less valuable.

Yep. I mean, that's what makes kind of, that's your moat, like kind of onboarding all these devices and all these functions. And that takes a long time to get it all integrated. You might do liquid chromatography, but you can't do mass spec. It's like, oh, Oh, okay, cool.

Right.

Right. So you can, you know, you might be able to do an ELISA or something, but then if you can't do, you know, PCR in the backend, so I think totally agree on that. Maybe we could shift gears and talk about, you know, examples of ECL. And I think one customer that you announced was going back to your alma mater, Carnegie Mellon, and it comes full circle, right? And he said, okay, That was really— when I saw that, I was like, oh, that's so awesome. You and Brian are going back to your alma mater and installing a huge site for them and doing a lot of work for them. And so maybe you want to talk about that, and we can also discuss other case studies of where ECL has really helped companies in drug development and beyond.

Yeah, absolutely. So for Carnegie Mellon, we're building them— so most of the way, the primary way that people use the ECL is through what we call our global cloud facility. Where it's a shared laboratory that all of our customers from, you know, 3-person startups to single-person like academic labs to giant biopharma companies can access from anywhere in the world. And it's sort of a shared resource that everyone has bandwidth on. But for large, for large individual customers, we also have this idea of what we call an on-premise facility where we will build and operate a cloud lab for an individual customer. And so that's what we're doing with Carnegie Mellon. We're building the world's first academic cloud lab that will be used by not just their, you know, faculty for doing research, but also for teaching classes, for collaborations, for startup companies that they're spinning out will have access to this facility. So it's really a tool for the entire university and not just for the biology and chemistry departments, but also for, you know, collaborations with robotics and computer science and computational biology. And one of the things that I think Carnegie Mellon has always done well, and one of the things that I'm most proud of is just this ability to really blend disciplines. And I think that this is going to be a sort of central resource that allows that to happen at an even faster time scale on the sciences side.

Really cool. I mean, yeah, I think that you have more universities on the horizon, I'm pretty sure of. And maybe do you have any, like, case studies or examples of existing customers in biotech and how that's been in terms of, like, you know, what you've enabled enabled? I think a big advantage is like if you can automate data generation, that's a prerequisite for like machine learning being useful in biology. And so I think that's kind of like, you know, in Citroën and a bunch of these companies, you kind of need to build your own infrastructure to even have the benefits of ML. But do you have any kind of examples of what automation has enabled in maybe medicine?

Yeah, so I mean, so some of the examples are very are very pedantic, which is just doing things like companies, and we also have some people in the academic realm doing this, so you can better predict purification conditions for a given compound. So, you know, you feed in or generate data for, you know, different classes of compounds, whether they be biologics or small molecules, and from that dataset, you can learn, okay, if I have another molecule that kind of looks like this, what's the right column? What's the right gradient? What's the right buffer system? So you can start to do some really interesting things to your point on the AI and ML side that are sort of like just next level abstractions on the system. We had, you know, so, but some of our customers are doing things that are much more pedantic where they're really just like transfer, like some of them, some of the bigger companies are just transferring methods that currently exist at, that they're currently running at, uh, at, you know, contract research organizations to the ECL. And the fact that they're able to just sort of like run those things in the background and can modify them, you know, not by like having to spin up another project and, you know, like have, you know, a half dozen conversations that they just like go into the, go into our software and it's like, okay, I'd actually like to see what this looks like. We use a different column or try a different gradient or, you know, run them, do a triple quad mass spec instead of, you know, some other method. So there's those sort of like, you know, very useful but pedantic examples. And then we have people who are using our, my background in graduate school was nucleic acid chemistry. So I'm particularly fond of companies that we have on the system that are using it for generating these novel nucleic acids or novel peptides where they're incorporating non-natural bases or non-natural amino acids into their oligomers and then QCing them and then testing them and sort of can build sort of like that full life cycle for those molecules into the software and the commands of the ECL. So we're trying to, and I'm trying to think of some other things like that we're, that we have, and I'm trying to tiptoe around confidentiality agreements that I've signed for some of this. But I mean, we also have things like, you know, we have CPG companies like testing, you know, testing over-the-counter like consumer products like on the system. So it really does span the range from some of the coolest cutting-edge startups that like doing things that are like new pharmaceutical modalities and things like that to like, you know, we've tested mouthwash on the system. So it's really, it really like runs the, runs the full like gamut. But I think that all of that is interesting because all of that is in the spirit of accelerating science, which I'm very passionate about. Absolutely.

I mean, if you can kind of empower more creative people even beyond drug development. And I think there's a big opportunity in like consumer products and industrial products to kind of use these tools to create like, you know, we're an investor in Revella, they're making like Rogaine 2.0. And so it's actually really cool. Evan's style is really good. And so maybe how do you think about, and you have these customers, you built out this platform, how do you think about scaling that platform? And has that into play with the business model. You know, there's opportunities like a marketplace potentially. Yeah, there might be network effects. Maybe you can touch upon that and how you think about e-sale over the next 10 years.

Yeah, absolutely. So I think that the, the highest sort of order bit to think about this is in terms of levels of abstraction. So if you think about other technology industries that have made rapid progress over like the last, you know, you know, half a, half a century, the thing that they have in common is that they've built layers of abstraction. So if you think about like the earliest like programming languages, like first you're writing machine code and then you have something like C where you're still like specifying addresses in memory and giving the microprocessor like specific instructions, but it's a little bit more than machine code. And now you have programming languages where there's not even like there's no even concept of being connected to a physical processor at the end of the day. And, that abstraction enables you to do increasingly powerful things on the system because the details kind of you can get glossed over. And, this is admittedly hyperbolic, but up until recently, whether or not something went into clinical trials could come down to, "Did I pipette into that well yet or not?" And so what we're doing right now, like sort of like the V1 of the ECL, is just the first layer of abstraction between the world of bits and the world of atoms that you can fully specify in software what you want to have happen in the lab. But that's just the first layer. There are many, many more layers that can be built on top of that. We certainly have things that we are, you know, currently working on toward that and things that we hope other people will build on for that. And you mentioned this idea of a marketplace, which I love. So if you're an individual scientist and you, like, design a new assay or generate a particularly valuable data set that you could give, you could basically sort of containerize that and have that as an application on the ECL that you can that you can monetize, that other people who are interested in that data or interested in that experiment or that assay can actually get access to that application on the ECL and then use that allows, again, it's in the spirit of moving science forward faster. The canonical thing that everyone says about science is that we stand on the shoulders of giants, which is certainly true, but really, if we want to move forward the fastest, we need to just stand on each other's shoulders. So the more that we can make data accessible and shareable and sort of quantifiable in this system through, you know, applications or a marketplace— there are many different ways that we're thinking about doing this— you can actually share, you know, not just data but also methods. And so ways of tying methods to data back to methods that you can then modify and generate novel datasets off that first one. I think that where all this is going is it's very exciting, especially as we're able to think more and more about these things conceptually and abstractly and not having to worry about the implementation details in the lab.

Absolutely. And then maybe to build on top of that, I don't know how much you want to disclose. I don't want to I'll let you just talk about it, but like in terms of what that enables for an individual partitioner of say services. And you know, what I'll say is like, you know, 20 years ago in biotech, there's been this massive move towards virtualization. But that means that a biotech company, when I start off, instead of buying all this lab space, I'm going to use CRL, Charles River, as a CRO.

Yeah.

Let them do the work. And then They do the work, and then maybe if the science plays out, they'll start the company and finance it, and then it gets started. Uh, and ECL is kind of the, maybe the final form of that, you know, 20-year trend. Uh, and on the backdrop of that, you know, all the CROs are consolidating, you know, something like 5, I think the CRO market, the top 5 CROs own like 70% of the market.

It's ridiculous.

The whole middle market, the whole, you know, boutique med chemist. Yeah, he, he's out of business, you know. And so how do you think about, uh, you know, ECL empowering those individuals, uh, to then, you know, help drug companies and other types of companies?

Yeah, so I, I think that— I think you can kind of— this is again overly simplistic, but you can kind of think of the CRO space as sort of like two— it's like a tale of two cities, right? You have the companies that really are all CROs are a combination of consultancy and experimentation. And the more you have ones that are really heavy on the consultancy side, or what you're paying them for is their expertise. And I think those people, we, you know, like the, you know, like the brilliant med chemists or the brilliant, you know, people who are like, you know, KOLs in, you know, DMPK assays or things like that, where they just, they are going to know more than any individual customer, and that's their value add. I think we can really empower those people because let's face it, like relative to like a consultancy, running a lab is a relatively low margin business. And most of those people ran labs because they had to, not because they wanted to. So if they can basically become meta CROs where they have the relationships with their customers and are designing the assays and then using ECL as the backends to generate the data that their customers need. To me, that's like wins all around. The place where it gets a little trickier to your early comments about Wuxi are with the CROs that are effectively just labor arbitrage plays. And if we can mitigate the value of that by using, you know, automation and, you know, 21st century, you know, practices in manufacturing and how to like do industrial engineering, which all of which we are doing and is sort of a secret passion of, maybe not so secret passion of mine, that I think that is sort of, if we can move the CRO market more toward this meta level of like consultancy where you have world experts who are helping advise you and, you know, help you direct your experiments as a company, whether you're a giant biopharma or a tiny startup, that is, you know, to everyone's benefit.

Absolutely. I think it would— you spent 10 years, ECL, building out all this really compelling infrastructure. And I think the next 10 years, it's growing that top business, but also building a whole new business model of like, you know, like a Shopify for drug development. And seriously, I like that.

That's very catchy.

And so I think you have to go do it. But you just look at the market dynamics, the market's— the timing is kind of right for that. And not only with the CRO market, but also with— there's a big need for onshoring biomanufacturing to the US due to COVID. And so on that point, and so, you know, a lot of your capabilities for ECL right now are focused on, you know, FACS, mass spec, maybe things that are more involved in, you know, maybe small molecules. I mean, I don't want to misframe it, But how do you think about expanding your menu of products? And maybe one day ECL can help companies with antibody discovery, automating gene therapy manufacturing, a whole host of other functions, and going up the value chain, so to speak, towards, from kind of general experiments into more and more customized experiments. Is that part of the purview, or do you want to kind of just monopolize kind of the more general experiments?

Making it so that you can get access to any instrument that you want is really like our first sort of— is the highest order bit for us right now. So the idea that, you know, so for example, you know, there are lots of different, you know, we don't have— there's no experiment like kinase assay on the ECL because there are, you know, there are lots of different ways to run a kinase assay and our job as we see it, is to give you access to the plate readers or the incubators or whatever you need to design the assay exactly how you want to run it, not to have this formulaic view of like, here's how, like if you want to run a kinase assay, you do it this way on the ECL. Having said that, I think that that level of having higher order functionality is something that we may get interested in, we may do some of that in the future. Some of that may be more of this app store that we talked about. I think that we do have an interest in, you know, right now our goal is sort of first order bit is getting everything that spans the range from cell and molecular biology through synthetic and analytical chemistry, so that there shouldn't be anything in that along that spectrum that you could do in a traditional lab that you can't do in the ECL. From there, we've already sort of been getting pushed a little bit by some of our customers to move more toward the inorganic chemistry and material science world. And that's, that's actually relatively low-hanging fruit for us just because the, you know, all the instruments are about the same size and all of our automation, you know, tricks and know-how and industrial engineering practices sort of feed very easily into that. But I think that there are, you know, this idea of allowing scientists to think less about, like, the, the implementation details in the lab and more about their high-level objectives is really important. And that's certainly the way that we would like to see this going, especially as you're seeing this, like, explosion in the number of different pharmaceutical modalities that companies are pursuing. I think that's really the, the direction that all of this is moving.

I totally agree. I think there's, uh, I think you have a lot of opportunities ahead of you. And so and just trying to figure out where the growth is and where you can kind of be first in class. So it's gonna be really, I think that the university path's gonna be really exciting, and that could be a good entry point towards expanding to a lot of other types of functions and whatnot. But maybe to kind of wrap it up and tie it all together, thoughts, and especially we have a lot of founders here, or potential founders. We've touched upon things around finding product-market fit, a little about team building, but do you have any kind of like lessons you've learned over the last decade? Maybe, you know, topics are like how to build a board, how to manage a board, how to set a culture, but any kind of specific things you think are worth sharing?

Yeah, I mean, I feel like I have, and I've developed over the last 10 or 11 years, plenty of scar tissue around all of those things. And so, I get, but just in general, I'm very loath to give advice. I feel like advice is just codified scar tissue. So, whenever I talk to people, so maybe this is a piece of advice, whether you're talking to an advisor or a board member or anyone in your network and you get advice, my default is always to ask, like, what's the story that led you to that piece of advice? Because we're scientists, right? Like, give me the raw data. I'll do my own analysis. But what I really want is the raw data. I want the stories that led you to say, like, you should never take money from this kind of investor because, like, the circumstances that led you to that conclusion may be different from the ones that, uh, maybe different from the ones that I'm in. And so I need to know, like, why you're saying that, not just, like, that that's the conclusion you came to. So my, my advice in general is always, like, get the stories, because the stories are invaluable. The advice that comes from those stories, you know, you can sort of take or leave, but I think getting those, those, like, getting that raw data, getting the stories from people who have been through it, either as investors or entrepreneurs or really anything, that's really where the value lies.

That's pretty powerful, really understand the stories of the advice you get. That's pretty, uh, that's actually pretty useful, uh, Serge. That actually very useful. I'm gonna start using that actually, as I hear, uh, because I get a lot of advice too as an investor. I'm always worried about like some other investor telling me something and I'm like, and then I go to sleep and then I— they, they bias me. So I'm always worried about biasing.

It's always, it's always funny. Like sometimes that advice is like, it's an N of 1, right? Like I had this really, you know, I, I'm just taking like a trivial example. Like, you know, we hired someone from, uh, from UCSF and they were complete like diva and basket case. So you should— so your conclusion from that might be like, don't hire anyone from UCSF. It's like, that's not the right conclusion from that. It's like You need to vet for like those personality things. So sometimes people have like valuable data but the wrong conclusion. And so that's why I feel like just getting the, getting the story that like the, and sometimes it's, you know, it's, uh, many, many stories that led them to that. And sometimes it's like an N of 1. So I think that all of those are, are useful to you as you sort of try to like synthesize that for your own company or project or team.

Awesome. Well, man, I had a great time talking, and we should, uh, you should let me— hit me up when you're in Berkeley. And, uh, you know, I'll be honest, I don't go to Palo Alto that often. Seriously, the joke is like SF is domestic travel and then Palo Alto is like international travel for me. So, uh, seriously, I love Berkeley, uh, but, uh, that's, that's pretty—

that's still pretty progressive. I feel like there are plenty of people who consider, like, if you have to cross a bridge, like, that's international travel.

I'm kind of like that. The people who live in Manhattan, you know, like, they still leave Manhattan for life. So I'm— that's kind of like me in Berkeley. I just have a sweet life here. And so I— if you hit me, we go to cheese— we go to cheese board or something, uh, when you're in town. But, uh, sounds great. Everybody in the audience, if you're a scientist, if you're running a company, and I would really recommend you check out Emerald Cloud Lab. And I'm sure they have a particular set of workflows that can really help you move faster. And so be sure to check out ECL. And I know for a fact there's gonna be a lot more announcements over the next few years. And so really inspiring work to not only build a great company, but like go through the ups and downs of it and like, you know, find product market fit. And like, you're still living the dream. You know, seriously, you're still living the dream of like what you and Brian set out to do 10, 15 years ago. So like, keep on living the dream. And I think we certainly— and we all look forward to, you know, ECL next 10 years, like what more you build. But DJ, I really appreciate taking the time, and we'll talk soon.

My pleasure. Always great talking with you. Thanks, everyone.

Thank you, everyone. Well, have a great day, everyone. We'll talk soon.

Bite. Bite.