Computer Science is So Hot Right Now | Grace Isford, Lux Capital

0:00I really think New York is poised to be the next hub for AI. And I think it's really simple. It's because the demand is here and the talent is here. A few like stats, which, you know, folks may not know. Right. 44 of Fortune 500 companies are headquartered here. There's also a lot of really good research talent here. So NYU Silver Lab. That's like the top AI lab that feeds a ton of folks to meta. It's led by Yom Lekhan, who is Meta's chief scientist. The top labs from Columbia, Cornell Tech, Princeton. They have a great NLP Institute with Dante Chen and Karthik Narasimhan, as well as just great international talent.

0:44Startups moving to the U.S., whether it's from Europe or Israel, new grads. And so like a few other stats and then I'll share a little bit of the Lux portfolio. Majority of the Lux AI portfolio is here. I think we have a pretty darn good portfolio. Hugging Face is headquartered here. Runway is headquartered here. Mosaic ML, which was acquired by Databricks, their research team is largely here. Osmo AI, that's ML for olfaction, if you haven't heard of it. They're based here. Modal, really cool serverless inference company based here in New York. They also have a Swedish presence. And so that's one of the drivers for them to be in New York.

1:20Together AI, AI cloud, they're crushing it, have a great go-to-market and eng office here. So we already have such a great and high-density group that it's kind of unsurprising to me that we've seen these great results already.

1:46Welcome to Sorcery. I'm Molly O'Shea, founder of Sorcery. Today, we have Grace Isford, partner at Lux Capital, and 2024 Forbes 30 Under 30 in VC. She joined Lux in 2022 and shortly after was named the youngest partner in the firm's history. Grace invests into companies innovating at the nexus of computational sciences. Since joining, she has sourced and led eight of the firm's investments, including Generative AI, Video Standout, Runway, now valued at$1.5 billion, Sakana AI, Maven AGI, and Langchain. Before Lux, Isford was a principal at Canvas Ventures, where she sourced$100 million in deployed capital and worked with 10 portfolio companies.

2:32This is a really fun conversation, and we go deep into AI, NYC, and future predictions. I hope you enjoy. Hi, Grace. Thanks for coming on. Thanks for having me. Super excited to be here. Well, it's a pleasure to have you. I've known you for pretty much my entire career. I think we met early on in a women's investing group in New York City or San Francisco or something. Definitely pre-pandemic. Yeah, 2018, 2019. It was definitely pre-COVID. Yeah, you've had an amazing career thus far. You've worked at Canvas. And since then, you've made some pretty incredible investments into companies like Runway, Lang Chang, Sakana AI, Together Reflex.

3:20Could you just share more on your investment strategy and how you got into this category of computational sciences? Totally. And thanks again for having me on. I've also been very impressed by everything that Sorcery has accomplished and honored to be one of your guests. It really does feel like we're coming full circle. And actually, I mean, I think you've followed my career for a while in some ways. So you kind of understand how the computational sciences all got started. But how I got into it was really goes back to when I worked at Handshake. Handshake is kind of like a LinkedIn for college students.

3:55I actually worked there as part of this thing called the Mayfield Fellowship Program, which I did at Stanford. And that's like a work study program where you get to actually work at an early stage startup. And when I worked at Handshake, we were a pretty small team, but we had a really sophisticated data and infrastructure stack. We were dealing with a lot of sensitive student data, helping curate that on these profiles that were private. and then in turn working with them to get, you know, both career centers and also kind of these enterprises on board to, for example, recruit a Stanford student to take your pick of a major corporation.

4:32And so I learned a ton there of, oh, machine learning is cool, but there's all these other systems and things that have to work like an API or like data reliability or like a data warehouse. I learned what that was. Right. And some of these early tools were coming out so that Like DDT had just come out. It used to be called a Fishtown Analytics. It was a consulting firm. Right. Census. Right. A lot of these kind of transform companies that are basically helping you work and move data around kind of your emergent data stack. So I really credit that time. And a colleague who I work with there is now major head of infrastructure at a major AI startup today.

5:10And so I credit that experience for really exciting me about the whole world of this kind of computational sciences, which is really a fancy way and a very nerdy way of saying cutting edge software. So that's all inner workings of how things work. So AI and machine learning, dev tools and open source, fintech infrastructure, blockchain infrastructure. How do you get kind of that end product in the hands of users? it can really go pretty wide reaching as far to like vertical software too and some of these vertical applications we've seen of AI. But what really gets me excited are these kind of complex technical problems to make sure that end user has like a super delightful experience.

5:52And so then, of course, Handshake led me to end up work part-time in venture at a firm called Canvas Ventures, where I ended up working full-time and I went down a bunch of rabbit holes with them. And one of the first, I think, was the API economy back in 2019, which if you Google me, you'll find some old medium posts out there of really trying to dive deep into the project. And for me, you know, a lot of the thesis work I do stems from an answer that I'm trying to find for myself. Right. I had a question of how do you categorize, you know, the API economy or maybe it's a solution to a problem solving.

6:29You know, why is there no good map of New York AI, given there's great talent here? And that often is how I kind of lead my own investing is just what are the questions I want to answer? How can I do that research and then hopefully share it in a way that's helpful for other people? I definitely remember the API economy. That database was great. Yeah, that was a good one. The air table was wonderful. And by the way, like big credit to you on computational sciences. I think I read that and I was like, damn, I should have came up with that. That sounds sophisticated. But, you know, it's not just enterprise software.

7:08It's a little bit more, I don't know, higher class nerd. Well, and I take that from Lux's book. And that's obviously where I currently work at Lux. And we are nerds, right? And we love things, we say, at the intersection of tech and science. And so, yes, that could be computational science. We also do stuff in the life sciences and also the physical sciences, which is a fancy way of saying, you know, a lot of space and hard tech and defense and manufacturing. And I think there's something really beautiful from a venture capitalist perspective of taking technical risk and backing scientists and backing technical problems, which in many ways, a lot of these computational science companies are of really figuring out, you know, the open source repository that's going to change how software is developed.

7:53or the machine learning inference technology that's the fastest throughput on the market. Those are the sorts of things that in many ways are pioneered by researchers and scientists that we can help bring to light with technology and with investment specifically at a faster stage than they could have without obviously venture funding. Going back to your research process. So you mentioned you start with a question and then you go down rabbit holes and that sort of thing. But could you draw that out a bit further? Like, I want to get into the mind of Grace. How do you break down these different industries and like pull on different threads and find where you want to ultimately go?

8:34So I think it definitely has to start from a kernel of a question, no matter what, right? Where it's, oh, you know, again, back to the API economy. What is the structure of an API company? What does that look like? Usually it's more questions than answers. Usually I don't know a majority of the answers, but it does start from an initiative for myself of like, oh, you know, this is a question I have. Usually then my next resort is reaching out to other people who I think are experts in the space or field. So that could be, you know, a major leader of go to market for an API company that could be going to reach out to a major research professor at a New York based kind of university here.

9:14And so going to really first principal sources of, okay, great. I'd love to hear exactly what you're seeing on the ground. What are you seeing that I'm not seeing? How would you explain this trend? And then try to ask a similar set of questions to several folks to have like a larger N of people to make sure these are not just isolated patterns I'm seeing. And then working from there to get information on who to speak to next. In many cases with the market maps, it's like, oh, well, if you're looking at, I don't know, healthcare API companies, you should talk to these 10 others. Or if you're talking about New York AI, you need to talk to these 10 other people.

9:49So it becomes a bit of a recursive flywheel where you usually have a main doc with maybe a framing and maybe my key thinking that will evolve over time as I keep updating it with a lot of this really cool, exciting answers and research. And then eventually I need to mold it and frame it. And sometimes that's easy and sometimes that's hard. And usually, for the most part, the market maps are intended to invest. in a company and trying to find, OK, where are the white space? Where are the biggest opportunity? Where are the biggest pain points? What sucks the most? That's usually where I'm trying to drive the map to.

10:24In certain cases, and I haven't published a lot of the ones I've done, frankly, because they are proprietary and sometimes investment intel, but is, OK, you know, talking to 20 people and realizing, oh, you know, this is totally not what I expected. Right. Or, oh, like the map of how this evolved, particularly in AI today is moving so quickly that what was a very great place to maybe be investing in 2021, 2022 is really not the place where there's a ton of opportunity today. So there's a lot of live factors you have to also take into account. And the timing of these articles, I think, and these market maps really do matter.

10:59Of course, I'll go back, you know, the API economy, I'll be referenced one more time, is like five years outdated now. But it's still a helpful basis if I run into an API company or if I have a thesis on, okay, how is this company disrupting the market? But it's not something that I'm constantly kind of pursuing and chasing in the same way I was five years ago. And then you kind of work and kind of compound all of these theses together where they're kind of in my stockpile of knowledge. And I can build on them should I want to. One that actually is a cool through line that has been kind helpful is I did a deep data infrastructure dive.

11:36This is again several years ago, which I think has been cool to see how that's evolved because a lot of those data leaders are now the same people adopting AI and LLMs and it's not like an AI team, it's just the data infrastructure tech team, right? And so a lot of the same problems we saw with that early data stack we're now seeing evolved or in parallel right with the AI stack today. So data reliability, like is the data reliable? The same question people are asking for large language models, right? Or is it working at scale observability, right? Is our, are we running into issues with our data pipelines?

12:12Oh, are we running into issues with reliability of our open AI API? And so in some ways, like history repeats itself in this industry. And so these market maps continue to be super helpful and these network and relationships of people who eventually start new companies or go elsewhere. Right. And you've already made a handful of investments at your time at Lux so far. So congratulations. I would love to dig into one in particular. So I'd say to frame this question out a little bit more, let's go through your sourcing strategy. Maybe how did you find this company? How did you find the other ones?

12:48But it's one particular company that I'm curious about because you know, it's not NYC AI, it's Japan. So you invested into Sakana. Like, how did you find this company and what kind of got you so excited about it? So the first thing I will say is we were not actually looking to have a Japanese investment. As much as there's a really cool full circle story on Sakana, because I actually lived for part of my childhood in Tokyo, Japan. And so it's been really special to now be going back there for board meetings. The one thing I will say, though, which does relate to New York, as I think Sakana is great proof, as is hugging face in our portfolio as well, that great talent is everywhere.

13:29And major metros are particularly great ways to attract awesome pockets of talent. And so I would expect to continue to see particularly great AI, but also just great engineering teams and talent in every great major metro. Right. We've seen New York developing, as we'll talk about probably later. Obviously, San Francisco has been great consistently. I think you've seen a lot of great stuff in London with the DeepMind team that kind of set foot and established there. Paris, between Mistral, Qtai, Hugging Face, large great engineering presences. And now Sakana, in many ways, is kind of like the DeepMind for Japan.

14:05And so that's really exciting and cool. And it's 100 million people. It's a huge economy where there's a lot of reasons why there could be a sovereign AI winner here. But to answer your question, we found Sakana a few different ways. The direct introduction came from another entrepreneur in the Lux portfolio who we had heard about because David Haw, one of the co-founders and the CEO of Sakana, is quite well known. He kind of pioneered creativity and machine learning research. He studied under Jeff Hinton. He then went to Google Brain after a long career at Goldman Sachs and worked under Jeff Dean at Google Brain and was a really well-known researcher, pretty viral researcher as well, for trying really creative and kind of weird contrarian research initiatives that often bear fruit.

14:53And so and the team really stood out here. And that's why we were like, OK, we'd love an introduction. We were trying to find a way in. Hey, we'll get right back to the conversation after a word from our sponsor. Sorcery is brought to you by Archer. I'm genuinely amazed at what Archer has been able to accomplish. Archer's goal is to transform urban travel, replacing 60 to 90 minute car commutes with estimated 10 to 20 minute electric air taxi flights. They are safe, sustainable, low noise, and cost competitive with ground transportation. Archer's Midnight is a piloted four passenger aircraft designed to perform rapid back-to-back flights with minimal charge time between flights.

15:31Learn more about how Archer is set to open up a new world of opportunity for passengers by providing safe and efficient access to people, places, and events across the communities they live. Visit Archer.com. David also did work and really admired the work of the Santa Fe Institute, which is this really cool research institution that does a lot of complexity research of how all the different things in our world kind of collide and come together, which it was another reason we were really interested to partner. And then the third reason, too, is kind of the Lux frontier tech, deep tech and defense tech angle.

16:07Right. Where we've worked kind of directly with a lot of great companies like Anderil at the Seed Round and Hadrian and many others. And kind of Japan's unique geopolitical position as a democratic U.S. ally. kind of is sitting in a really unique position to work with the U.S. in an exciting kind of AI strategy. So that's all the reasons why it was interesting. The last thing I'll say is, in addition to David, the other two co-founders are incredible of Sakana. So one is Lyon Jones. He was one of the authors of the Transformer paper. He is not Japanese. He's actually Welsh by background. I should get that perfectly correct.

16:47And he moved from the Bay Area to Tokyo to the Google office there and just kind of fell in love with it. And then Ren Ito, the COO, he actually spent time in Prime Minister Abe's administration. So he has Japanese background and worked closely there. He also was CEO of Mercari Europe and has a lot of that more commercial experience to complement David and Lyon. So it's a really exciting and cool one. I think to answer the sourcing question, some of our best investments come from kind of, I don't want to say random, but but like kind of serendipity, quite literally, where we're kind of creating our own luck of great.

17:29We have a great network and I think reputation and network is the best currency. is someone thinks of Lux or thinks of introducing Lux to a great entrepreneur, whether that's our own founder in our own portfolio or, you know, another friend or someone who gives them to me, that's the highest praise and the most important thing to do. So I've gotten I've done deals from, you know, close friends who are also co-investors and senior investors and angels. I've done cold outreach where I've read a blog post that a founder has written and I've said, OK, wow, it's super interesting. I have a thesis here.

18:02I have a market map here. I'd love to talk about it. I've also just straight out hustled. Right. And I hear a deal is happening and I'll reach out directly even when we know the company is actively raising. And then a lot of it is more thematic, too, of, hey, you're a cool founder or you look like you could be a cool founder. Often this is pre when the company is getting started. I love to just brainstorm with you. Right. And talk about your idea, how you're thinking about it. Talk about Lux's portfolio investments. And so we do really try to go early. And we also try to think about what our right to win is, whether it's a company where we have a lot of touch points with the founder or we have a lot of shared kind of thematic combinations.

18:41And so what stages are you coming in at? Usually seed series A. I would say the highest frequency of my investments at Lux have been seed. we've come we can come in and this is a little bit of a pun but it's true we can come in from your first 100k to your last 100 million which so we are a true multi-stage fund what that means in practice though is we typically write our first check to you at the pre-seed to series b and then those leader checks are for a company like hugging face andrel who is a bit more mature in our portfolio where we want to continue to grow and maintain our ownership and our winners And so that's where you kind of see that concentration of those larger checks come in.

19:24Secondary question, but I just have to ask, Grace, Hugging Face, how do they make money? What is their business model? Hugging Face is an incredible company. For those who are not familiar with it, I think most people probably are at this point. It's like a GitHub for machine learning, right? And so the most powerful thing that they have and they will continue to have and what really predicated our original investment was this incredible community. their GitHub repo that took advantage really of the key moment of the transformer paper unlock productizing that and then kind of becoming in many ways like the the toll booth of AI in some way and so that kind of relates a little bit to the business model um they monetize in a few ways they have kind of like a freemium model and that's like you know live on their website where you could get a pro account so basically it's just like access to the public hub but with a lot of extra nice things.

20:16So things like higher rate limits or extra access controls or things like that. So for a power user, it's a small fee to pay for a lot of value of accessing and working with a lot of these open source models in a sandbox. They have rev share partnerships and they have enterprise partnerships. But the company is doing awesome and I think is probably monetizing in an unconventional way, but in a way that's doing really well. And I've been super impressed by what they've done. Great. So I know that the portfolio is very technical. You guys invest into sciences. It's great. I loved that point in all different aspects.

20:57How does the team then evaluate each company on a technical basis? Do you work with consultants? Do each of you kind of go in and take apart the stack and understand what works, what doesn't work? How do you evaluate each of the companies? So at Basis, Lux team is already highly technical, right? As you kind of hinted at. So I believe we have at least two PhDs on our team. We have at least five or six masters. And then on top of that, we have multiple just bachelors of engineering as well on top of that. So I should know the number, but I think we have at least like 20 different degrees of various tenure in the sciences.

21:36And so I think that I don't think that lightly because I think that's that's pretty amazing. And some of our team members have done research for their PhDs, et cetera. I've done research and most of us have coded in some capacity as well. So I think that there's something to be said for just that from the investment team perspective and the network we have as a result of that. In terms of how we diligence, it does depend a lot on the company, Right. So the Lux team is a generalist team. So, you know, I am voting and working with companies across sectors, even though I have a passion and excitement for computational sciences.

22:12But my colleague, David, who's wonderful, you should meet. He spends a lot of his time in the bio world and he has a deep expertise. He works at the Broad Institute and did research there and has a lot of technical expertise there. And so depending on the deal, right, say maybe it was a bio or a bio AI deal, I would take the meeting with David or we would take the meeting together. Right. And so we're already compounding our effect. And the Lux team is of much of a team effort. So it's unusual for a company to come into partnership without having met multiple members of the team. Second layer is, you know, assuming it passes team sniff test is we have an excellent network.

22:50Right. In part, it's from our existing portfolio of investments. In part, it's people who we've known or met through the years. And in part, it may just be direct outreach. I think actually in part of the New York work I've done, I'm really happy to see the network that I've built right in New York myself. Right. So it's pretty easy for me to get in touch with a senior engineering leader of a very niche area because I have enough nodes into certain organizations. And so I think it's the key advice I guess I would give to someone diligently in tech is know the space generally. Right. Don't go try to be evaluating technical risk area you have no idea about.

23:28And if you really have no idea, get up to speed first and try to at least understand the market and do your own research. Assuming you are up to speed and excited about it, make sure you're building your network even in the downtime. Right. Of course, you want to pick up the phone and call that person on the day where you need to close the deal. But it's much more useful if you invited them to dinners, you've gotten to know them, you got coffee with them. And then, boom, when I need that diligence call, you know, they're going to pick up the call and they're going to help. And that's been really fruitful for me in terms of both diligencing, but also, you know, converting potential customers for, you know, prospective portfolio companies.

24:06And then it's also just a little bit of pattern matching and seeing investments over time. Meaning, you know, if I've seen 10 different companies doing, you know, enterprise AI agents, I have a bit of a perspective on what stands out in that space versus not. And so I do think it's a little bit of are you in the mix and seeing things that are happening right now? If I had only seen one of those 10 pitches, I may have been much more compelled versus having seen all 10 of them. You know, the 10th one is a little bit less interesting. So it is also about making sure you have good coverage of everything that's happening in the technical landscape and as much coverage as you can on the frontier things that are happening and changing and the new models that are coming out.

24:50Excellent. Super thorough. Obviously, you guys are a serious shop and it definitely shows. I want to expand on New York for a little bit. So you went viral for this market map that you made for AI and, you know, just kind of cementing New York as a figure in the tech world. There's always a debate. We love the debate. It's so much fun. But so, you know, like NYC is known for fintech and health tech, but I'd love for you to just break down the NYC AI angle and bring us all up to speed, what is the state of the market there? Well, I'm still hoping to get you back in New York, Molly. So I'm optimistic that maybe the West Coast has not went out on you.

25:37But maybe, you know, after this podcast, you'll be like, you know what, Grace convinced me I'm packing up my bags back to New York City. But I really think New York is poised to be the next hub for AI. And I think it's really simple. It's because the demand is here and the talent is here. A few like stats, which, you know, folks may not know. Right. 44 of Fortune 500 companies are headquartered here. It is a center of major industries, like you mentioned, not just, you know, financial services, but also media, fashion, et cetera, health care. There's also a lot of really good research talent here.