Radical AI’s Joseph Krause: The Scientist Building The ‘Waymo’ Lab For New Materials

0:00Joseph Krause:We really think we're going to be one of the most important companies in the world. When we put a civilization on Mars, the habitat that they're living in will be made with radical materials. That is what we believe. You might have used AI to write an email or generate a simple image. But a startup called Radical AI is trying to do something much more ambitious. Their goal? To revolutionize the scientific process itself. CEO and co-founder Joseph Krause says Radical AI is creating self-driving science labs. Humans work in cereal. I make hypothesis, I run experiment, I test results, I learn from results, I redo the whole thing.

0:36Joseph Krause:Our lab is doing all of those things simultaneously. Today on the podcast, we're going to talk about why Radical AI is developing new metals right here in New York. How Joseph cold emailed his way into a$55 million funding round to get started. And why Radical is more like a Waymo self-driving car than an AI science lab that might just be more like hands-free driving. Plus, why you can't buy your way to success with a billion dollars, at least yet. I'm Alex Conrad, founder and editor of Upstarts Media. And this is the Upstarts Podcast, our weekly show about startup founders who punch above their weight to take on the status quo.

1:10Joseph Krause:Joseph, thanks for joining the show. Alex, thanks for having me, man. Excited to be here. This episode is brought to you by Rippling AI, the only AI built to give you full visibility into your startup and the ability to take action across every department. So Joseph, what is Radical AI? Radical AI is a next generation material science company where we use AI and autonomy to change the way we do the scientific process, where we really move from what is a human-driven process today to an AI and autonomy-driven process of the future that fundamentally removes materials as the biggest bottleneck to our most important industries.

1:45Joseph Krause:That's what the company is, and that's what we're working on. That's awesome. And you just raised $55 million. You're opening a bigger facility in Brooklyn Navy Yard soon, right? Yes, that's correct. So we raised$55 million in our seed round last year and have been off into the races with that. In the new facility, we're taking a whole building, which was previously called Building 20 in the Brooklyn Navy Yard, 45 ,000 usable square feet. And what's really exciting about this facility is it allows us to, one, grow the team and expand operations, but to build a multitude of material systems there.

2:14Joseph Krause:So not a single material system like our current facility has today, but actually material systems that span different industries and different material markets. And we'll get into why that's really important, but that's an exciting spot in that facility. What are the devices like at a big picture that are creating these materials? Yeah, absolutely. So at a big picture, you know, first, let's take one step back. What does a material scientist do? So a material scientist, when they're working on a research problem, will start with hypothesis generation, right? So they'll sit down, they're going to identify this research problem.

2:45Joseph Krause:They're going to read a bunch of scientific publications. They're going to get new ideas and they're going to make a set of hypotheses on what they want to go make. They're going to go into the research lab. They're going to synthesize these materials, which means to actually create the physical material. They're going to characterize the material. So, hey, what did I make and what does it look like? Is the structure what I thought it was going to be? Did it come out how I thought it was going to come out, for lack of a better phrase? And then they're going to test the material. So, does it have the performance properties I want and was expecting when I made the hypothesis?

3:17Joseph Krause:After they do all this work, they're going to analyze that data and then they're going to go back in their brain and think about a new hypothesis on how to what we call iterate on those experimental results. Now come into our facility. At Radical AI, that entire process that I just explained to you is done autonomously. And all of the hypothesis generation, data analysis, and what we call active learning is done with artificial intelligence. And so you move from this serial-based process that a human scientist must do to this parallel-based process where AI and autonomy can do that scientific experimentation many times over for a material system.

3:59Joseph Krause:So that's what's in the lab. Now, for our current facility, we call this System 1. And System 1, Alex, was kind of our beachhead into building self-driving labs and the technology I just described to you. It is in a field that is called structural metals. And these structural metals make up a lot of the big industries we always like to talk about. Aerospace, automotive, nuclear, defense. They are what actually form the components and the products that go inside those systems. And so that's the research lab that we started. And all of our tools are built around the structural metals area today.

4:36Okay. So these materials are alloys that are going to be metals that are used in those various industries. That's correct. Exactly right.

4:44Joseph Krause:So you take a jet turbine, for example, alloys go inside this jet turbine that allow it to operate at extreme temperature and for very long cycles. The alloys that we're working on have an opportunity to make it into those new jet turbines of the future. Is it just metals or is metals just where you're starting? Like, are these materials here, like this composite wood that I see in front of me or this plastic here, like, would those be just less valuable materials or how should we distinguish? Not less valuable. I would just say different market, right? Different end product. And so even inside the applications we're talking today, of course, they have other material systems that are also being used, things like ceramics and composites, or maybe even into polymers.

5:24Joseph Krause:The thing that we really believe is first demonstrating the way to do science this way, right? If you really think about this, we are disrupting a multi-hundred-year-old industry in the scientific process and in materials, 150-year-old industry. And so we are approaching the discovery problem and then the ability to take the results from that discovery and push them to scalability in a different way than the industry does. So the first thing we wanted to do with the lab was really demonstrate, hey, this is what it means to run self-driving lab architecture. Here are the results. Here's how it works.

6:02Joseph Krause:Here's how much throughput you can do. That's different than the process today. Now, as we kind of build out that system, we want to scale to new material systems outside of alloys. Those are some of the other systems that will go inside Building 20 in our new HQ in the Brooklyn Navy Yard. So we're now at that expansion point. So growing up as a kid, were you obsessed with metal alloys and materials? And you're like, someday I will create a new metal. Or where Where does this come from? I was not. I didn't grow up a metal geek. I can't admit that. So I think I always knew I wanted to be an engineer.

6:37Joseph Krause:From the earliest I can remember, I used to go to family parties and be like, I'm going to be an engineer, whatever that meant to me at the time. And when I went to undergrad, I was going to be a chemical engineer. That was the best engineering that I knew. I thought it was really exciting. And I took an introduction to nanoscience class when I was in my undergraduate. What was so cool about this class is nanoscience is all about the science of the very small at the nanometer scale. All of the interesting end applications that we see today, particularly in like electronics and semiconductors, or even in understanding something like the performance of a metal in a jet turbine, has characteristics and properties that come from that nanoscale.

7:18Joseph Krause:And so nanoengineering started to be this new term where it was a blend of material science, It was a blend of physics. It was a blend of chemistry. It was all the engineering coming in to make those things into real products. It was just this meshing of worlds that I was so excited about. So did that in my undergraduate nanoscience. And then I did a PhD work, did my PhD work in nanoengineering. And that's kind of how I moved into, I want to apply the meshing of these scientific fields into making new products. The way we got to metals was it was one of the hardest problems that we could solve.

7:50Joseph Krause:when we looked at the material science base and wanted to build this technology, we knew that if you could do it in metals, we think we can do it in almost any other field in the world. Probably the only other field that is equally as challenging is semiconductor research and some of the materials that go inside of those. We're also taking on that problem and we can talk about that, but we knew it was a place where no one had built automated solutions yet and it could really distinguish how effective our approach was to the process. But you did do a bit of investing and sort of a walkabout in the startup ecosystem before starting Radical.

8:24Was that trying to develop your point of view as a future founder or what was the strategy behind that?

8:29Joseph Krause:I was in graduate school, I was in a PhD and I was doing a fellowship at the US Army Research Lab, which is an amazing institution. It is this research lab that really does fundamental research, but for the US Army and Defense Department. So they have tied objectives to what they're doing with that research. the problem is still fundamental in nature right early on the technology readiness level scale as they typically call it at dow and for me i had this constant want to how do i get this stuff to the real world how do i get these things into commercial products and i know how to do that so i cold emailed like a hundred vcs here in new york i wanted to move to new york and one of them was Kevin Ryan.

9:12Joseph Krause:Kevin met with me and I pitched Kevin on, hey, Kevin, if you're not investing in material science, you're not investing in the future. And he was like, okay, I've been doing this for 35 years. I've started multi-billion dollar companies. I don't know how right you are on that thesis, but why don't you come to New York? I'll give you a six months internship. Either it'll be your new career field or you'll go back and finish your PhD. I said, deal. Packed my stuff up, moved up to New York a week later and started investing in material science. what we did and what I learned from that process was one what makes a good company at least from the investing seat right what company succeeded what company struggled how did they approach hiring how did they approach go to market how did they think about and when to think about scaling the team and after we had this new idea we were looking into physical world and where is AI going to make an impact and naturally my materials background I said what about material science and my two other co-founders, one coming from the materials realm and one coming from the software realm.

10:10Joseph Krause:We just all saw that and attacked it. And so that's kind of how Radical came to be. And so to answer your question directly, didn't go right to the founder journey, more walked my way up through seeing it from the investor lens and then took a shot in the founder seat. It was interesting to see Radical kind of come out. There was chatter like, hey, is there this thing being incubated out of Alley Corp? Is this part of Alley Corp? Is it a new lab? and this is in the really kind of heady moment of 2024, right? So ChatGBT has come out, AI applications are starting to excite people in different areas.

10:43Was there a clear correlation there between AI explosion and this is the time for you guys to be doing Radical?

10:49Joseph Krause:Yes, I think there was, particularly around our frustration with where we saw companies being formed in the AI wave. Exactly as you just said, this wave was taking off, everyone was investing, there was pressure to deploy dollars in really exciting companies that are going to be the big companies of the future. And me and one of my other co-founders, Jorge Calendras, who was also at Alicorp at the time, were so frustrated. We were just like, how many more recruiting pitches can we get to like redefine the recruiting industry, right? I'm sure it's a good problem. Nothing against those businesses, but we were looking for something more.

11:24Joseph Krause:This is where we started thinking about materials. We thought about robotics. We thought about manufacturing and like physical industry. and we netted that if you really want to leave a fundamental impact on humanity, then materials are one of the best ways to do that. Because regardless of what industry you care about, automotive and aerospace, manufacturing and defense, climate, energy, semiconductors, electronics, the most important industries in the world are all direct result from materials R &D. So to us, it felt like the perfect blend where you have this technology that is completely changing the paradigm and discovery and you have this field that's 100 something years old and hasn't had it happen yet.

12:01Joseph Krause:That's the place that we should attack. Let's talk about that because zooming out, I could almost imagine people saying, well, isn't it arrogant to think that this industry has been the way it is for 150 years and these new young guys are going to transform it in just a handful of years? You know, good luck, buddy. Like, why hasn't that worked before? There is a very important thing with self-driving labs that we think unlocks materials today versus materials of 10 years ago. and that is called concurrent engineering. And I'm going to talk about that. But the first thing worth mentioning is persona.

12:36Joseph Krause:We don't give a crap what the industry thinks. And more importantly, we really believe in first principle analysis. And I think that ability to question the way the industry does things is where the best innovations typically come from. You know, it's like innovators dilemma, but on steroids, because you have an industry that's also baked in with protection from the difficulties of starting something new. So not only are they like not innovating a lot, but they also haven't felt a need to over the last 25, 50 years. I mean, even if you correlate that to like the semiconductor industry, yeah, they might have some innovators dilemma, but they still have an immense pressure from the market to innovate quickly or get beat by another competitor coming up, whether that's at the foundry level or the ASIC level.

13:18Joseph Krause:So there was this kind of dual thing going on materials. Now, this is really important to understand. Materials of the past have taken so long to develop that people that engineer products do not consider new materials when they do so. When I'm building a new rocket, I look at materials that are 10, 20 years old, have been what is called qualified and have reliable supply chains that I can tap into. All of my design software and my design tooling has materials that have qualified supply chains and can be tapped into. Because you don't want to be doing a big project on a house of cards that then it turns out this material is not what you hoped and everything falls apart.

13:58Joseph Krause:That's exactly correct. And then the time. If I'm going to go make a new material and it's going to take me 10 years realistically or 15 years realistically to get there, I'm going to be on product iteration 15 by the time I get there. Certainly when you think about a company like SpaceX who moves incredibly fast and has built systems from the ground up. So SpaceX is actually one of the ones that pioneered this. a gentleman by the name of Charles Kuhlman, who is the VP of engineering for materials at both SpaceX and Tesla. They needed a new alloy for something inside a SpaceX component. And rather than go to the market, and I'm sure they probably tried this.

14:32Joseph Krause:I wasn't there, so I don't know. They actually engineered the alloy themselves. It's called SX500. No one really knows what it is except for a few people, but they vertically integrated down to the materials. And they did this concurrent engineering where as I'm developing my product, I'm developing a new material. And that ability to do that has really opened the aperture on what you can do with material science. I talked to you earlier about so many problems being linked to a lack of materials discovery. Well, if I build a new world where I can now do materials research at the speed I can do product research and development, now anyone can build a custom material for their application.

15:13Joseph Krause:That is really impactful and very powerful. So that moment, That ability to do concurrent engineering with these systems did not exist two, three, four years ago. We didn't have the AI systems. We didn't have the autonomous systems. That technology was not ready like it is today. But there was a wave of startups that pitched cloud labs and automated labs for both drug discovery but also for materials. I visited several of these in South San Francisco like 10 years ago. And those companies, which VCs were really hyping 10 years ago, some of them still exist. I did a little research. They're not gone, but they're certainly not getting the excitement and the buzz that they did back then.

15:52What was missing in the last wave that you guys have cracked?

15:56Joseph Krause:Great question. Particularly automation versus self-driving. This is a very important concept to highlight. I always like to use a simple analogy for people, hands-free driving versus Waymo. In hands-free driving, I'm still responsible for driving the car. I have to know where I'm going. I have to know when I need to make a left-hand turn. If a pedestrian runs into the crosswalk, I have to slam on the brakes. Yes, I can maybe take my hands off the wheel or my throttle control is set and controlled by the car, but I am still the one driving the motor vehicle. Now think about a Waymo where none of that is true.

16:30Joseph Krause:I put in my destination. I get in the back seat. I can go to sleep if I want, laying down on the bench and end up at my destination. I don't really care if it took five left-hand turns to make a circle because I'm asleep in the back. I just want to end up with where I'm going. We are building the Waymo of materials research. And in the past companies who have built, particularly on automated labs, it was really the hands-free. It was, we can do experiments for you and we can do a lot of them, but we can do the same experiment over and over where we can run just a high throughput number, not learning from all that data, building an active learning loop, and then having an AI hypothesize new materials from the results of that.

17:11Joseph Krause:That whole back part of that system did not exist. So the aperture was kind of too narrow. It was like, we can do this at a much higher volume, but you're going to still need to spend a lot of time telling us what to do versus figure it out. You know, it was kind of like, we can assist your current process today by making it easier and faster. The same way hands-free driving is, hey, we can assist your current driving today by making it easier and more intuitive, but you still need to pay attention and run the research. That is what those cloud labs used to offer. Today, we're like, no, no, anyone can run the research.

17:46Joseph Krause:I mean, you, Alex, could run research on our platform if you want. You don't even need a materials background. Now, the materials background will help you refine what you probably should work on, but you can do that. And so it's just an entirely different approach from automated versus self-driving. Okay, when we think about the upstarts, metal swag for season three. We'll get back to you. Absolutely. These days, you can chat with AI about almost any business problem. Rippling AI is built to actually solve them. That's because Rippling AI is built on your live workforce data that gives you full visibility into your startup and the ability to take action across every department.

18:22Say you're planning your next few hires. Just ask Rippling AI, what would it cost to add two more engineers this quarter? and you'll instantly see a breakdown of comp ranges by level and location. So you can make the smartest hiring decision based on where your team has gaps and where the talent is. But it doesn't stop there. Rippling AI can draft the job recs, route them for approval, and kick off the hiring workflow. All you have to do is tap confirm and then get back to building. Don't settle for AI that's all talk. Head to rippling.ai slash upstarts and get AI that turns insights into action.

18:53That's r-i-p-p-l-i-n-g dot a-i slash upstarts

18:59Joseph Krause:Sign up for exclusive access today Okay, so as you're setting up Radical Do people think this is a crazy idea? It's ahead of its time It's too good to be true What's the reaction from the market? Pretty crazy Way too capex intensive Self-driving labs will never work There was already all these cloud labs in the past We don't get what's different We don't see the AI wave catching up As fast as you're saying it's going to Spoiler alert, it did and it went way faster. So it was just a lot of resistance on, are you really going to use AI in a capex intensive, industrial based legacy industry like materials?

19:36Joseph Krause:Yes, we know materials are important, but that entire process feels very scary and very complicated to do. That was the exact reason we thought it was important to do. So that was kind of what the reaction was when we first went out the market. Obviously, a lot of startups get doubted by the incumbents, But did that make it hard to raise or to hire? Or when you think about sort of the cold start problem you guys had to overcome, where was the biggest challenge? So there are two. The first is in the infrastructure. Tooling is a really good example. And this has started to change now. We have some really good partners and companies that we've worked with that want to be at the forefront of self-driving labs.

20:10Joseph Krause:But two and a half, three years ago, when we went out and talked to tool providers, so material science tool providers, they weren't really in the self-driving labs. We have heard about automated labs. Yes, we know what they are. We don't really see automated labs being a big part of our business per se. We've heard this wave come up in the past and it just hasn't moved forward like we thought it was going to. And so a lot of the stuff that we had to build, you know, a lot of the APIs or SDKs, the PLCs that we actually connect our software stack to, we had to custom build, right? Or we had to pull teeth with some of the tool providers to actually enable us to have access to their system so that we could control it.

20:49Joseph Krause:Today, entirely changed. Self-driving labs are here. Every company is building them. The US government has invested aggressively in those. Other world governments, Canada, Switzerland, Germany, all have investments in self-driving lab architecture. So now we feel a lot more responsiveness and actually field some inbound from tool providers that want to build automated solutions and make their tools a part of self-driving lab architecture. So it's been an entire change there. So tooling was the first. The second was the research teams at large companies. So if you go to the fortune 500 and you look at their materials r &d teams spend a lot of money But they do materials in a very Standard way the process I described to you earlier this serial based approach where we move through this process slowly It takes 10 years to do a new discovery Because of that a lot of their work is focused on what we call optimization I'm, not going to work on creating a brand new material if we're 15 years away from using it how am I going to drive value to my organization?

21:51Joseph Krause:You got a lot of these projects and there's a lot of this research that was 1%, 5%, right? 10 % gain on existing materials we already have. We don't want to do that. What we want to enable is creating the markets of the future, right? We want to enable the materials that do not exist today, that change performance and enable new technologies to form from that. That is really what this system is built to do. And so we've kind of been working with the corporate R &D teams who want to use CloudLabs on showing them that you no longer need to just optimize your existing stack. You can actually create a new one as well.