Meet Jenny Xiao, Partner at Leonis Capital

Jenny Xiao is a Partner at Leonis Capital, a research-driven VC fund focused on seed and pre-seed AI companies. Before Leonis, she was an early employee at OpenAI and a Ph.D. researcher at Columbia. She left OpenAI just a week after ChatGPT took off to join forces with her partner, Jay Zhao, to build a research-driven fund backing AI-native companies. Jenny’s research has been featured in Nature, the NYT, the WSJ, and Bloomberg, and her approach is the subject of a Harvard Business School case study. She is also a Forbes 30 Under 30 honoree and a frequent speaker at conferences such as Fortune Brainstorm Tech.

Judy: Let’s start with your background: You were at OpenAI before joining Leonis Capital.  How did you go from being on the operator side to a VC fund?

Jenny: It’s interesting because I never imagined myself becoming a VC. I was deep in academia, working on a PhD at Columbia, and ended up leaving that program to join OpenAI.  My parents were definitely not thrilled. But looking back, it was absolutely the right decision.

OpenAI was very much a startup when I joined (only about 100 people) so I got firsthand exposure to that environment. Plus, our team worked closely with other startups using OpenAI’s APIs, which gave me even more insight into the ecosystem.

I’ve always had an entrepreneurial streak. In college, I started a small AI consulting firm with a couple of friends, and OpenAI actually became one of our clients, which is how I got connected with them in the first place. I’ve never been someone who stays at a company for too long. I tend to think in terms of launchpads: “What’s the next thing this opportunity can lead to?”

When ChatGPT took off, I saw a window to start something new. I was torn between launching a company or starting a fund. What tipped me toward VC was my then-boyfriend (now husband) who was starting his second company. My mom gave me some great advice: “You can both be entrepreneurial, but you can’t both be seed-stage founders.” That really stuck with me and here we are.

Judy: Who helped you break into venture? How did you meet your co-founder Jay?

Jenny: I was introduced to Jay through a friend of mine who works on the family office side. At the time, I was leaving OpenAI and exploring what I wanted to do next. My friend said, “You have to meet Jay,” and that introduction turned out to be one of the most important ones I’ve had.

We hit it off immediately. Jay had already launched his first fund about a year before we met, and coincidentally, a couple of my friends who were early employees at OpenAI and Google DeepMind, were LPs in that fund.  There was already a layer of trust and familiarity, which made the partnership feel natural; we were also aligned on vision, values, and the kind of firm we wanted to build.  Jay was the perfect co-founder to start this journey with.

Judy: What differentiates Leonis Capital from other AI-focused VC firms?

Jenny: I think what really sets us apart is how deep we go, both technically and strategically.  We aim to be the most knowledgeable and thoughtful investor on the cap table and Founders often tell us that our diligence process is unlike anything they’ve experienced. 

We ask questions that go far beyond the typical “What’s your revenue?” or “What’s your go-to-market strategy?” For example, we recently backed a company doing RNA sequencing using state space models. We dug into why they chose that architecture over transformers, what infrastructure gaps they’re facing, how many GPUs they’ll need to scale - it’s very technical, and that’s intentional. We want to understand the core of what they’re building.

Another big differentiator is our research. We publish deep technical pieces that founders actually read. One of our pieces on MCP tech stacks became the most widely circulated write-up on that topic in technical circles. When we reach out to a founder, they often already know who we are and what we stand for. That kind of credibility is hard to build, and it’s something we’ve been very intentional about.

We also avoid the hype. We’re not chasing thin wrappers around foundation models, those middle-layer companies that sit between the API and the user. They might grow fast, but they’re often competing directly with the model providers and struggling with margins. Instead, we look for companies that go deep into verticals, such as in bioinformatics, construction, healthcare - places where domain expertise matters and where AI can create real defensible value.

Our team has depth, credibility, and a very clear thesis. We’re not trying to be everywhere; we’re trying to be exactly where we can add the most value.

Judy: Can you share examples of companies you’re excited about?

Jenny: One company I’m really excited about is Kepler AI. They’re tackling a huge problem in bioinformatics and bridging the gap between wet lab scientists and computational tools. The CEO is a bioinformatician from Stanford, and the CTO used to lead AI research at Databricks, so they bring a powerful combination of domain expertise and technical depth.

What makes Kepler stand out is how personal the problem feels to me. My mom is a cardiology professor who works with large biological datasets but doesn’t code. I spent a couple of summers helping her run analyses: setting up Docker containers, spinning up GCP clusters, and translating her research needs into code. It was clear to me how inefficient and frustrating that process can be. Kepler is building AI tools to automate that workflow, and I immediately understood the value and the market. It’s not just a biotech problem, it’s an AI infrastructure challenge, and they’re solving it in a way that’s deeply verticalized and technically sound.

Another company I’m excited about is QualGent. They’re using OpenAI’s operator model to power QA testing. What’s interesting is that I had tried using the same model to play Age of Empires (yes, I’m a huge gamer) and when I saw their approach, I instantly got it. They were the first to apply this model to QA testing, and it’s exactly the kind of technical innovation I love to see. It’s niche, it’s clever, and it’s grounded in real product needs.

And then there’s Motion, founded by Harry Qi. He’s one of the most product-obsessed founders I’ve ever met. We didn’t meet him through a pitch deck - we booked a demo as users. Instead of trying to sell us something, he wanted to understand our productivity barriers and how Motion could solve them. That level of customer obsession shows up in everything they do: from product design to go-to-market. They’ve scaled aggressively and are now valued at over $550 million.

What ties all these companies together is that they’re deeply vertical, technically differentiated, and led by founders who are extreme in their focus. That’s the kind of energy we look for at Leonis.

Judy: What kind of founders resonate most with you?

Jenny: I tend to gravitate toward founders who are extreme in some way, not necessarily well-rounded, but exceptional in one or two dimensions. I’m not looking for someone who’s a 4 out of 5 across the board. I’d rather back someone who’s a 7 out of 5 in one area and maybe a 2 in another. That kind of intensity usually translates into outsized outcomes.

I also look for founder-market fit. In AI especially, the barrier to starting a company is low, so you’ll often see 10 or 20 teams going after the same idea. What I want to know is: Why you? Why are you uniquely positioned to win in this space? I want to see a deep connection to the problem, whether it’s through personal experience, technical expertise, or a unique insight.

And finally, I look for clarity of thought. Founders who can articulate their vision crisply, who know exactly what they’re building and why.  Those are the people I want to back.

Judy: How do you support founders post-investment?

Jenny: The reality is, founders drive 99% of the success. Our role is to contribute that remaining 1%, but we try to make that 1% really count.

For us, that means being strategic partners, not just capital providers. We help shape the company’s positioning in the market, think through competitive dynamics, and offer perspective on key decisions, especially around product direction, hiring, and fundraising. We’re not in the weeds day-to-day, but we’re always available to help founders zoom out and see the bigger picture.

One of the most valuable things we do is connect founders with others who’ve faced similar challenges. Whether it’s someone who’s scaled a team in a similar vertical, navigated a tough fundraising environment, or built out a technical infrastructure, we make those intros. Peer learning is incredibly powerful, and we try to facilitate that as much as possible.

We also help with hiring, especially for technical roles. Because of our deep network in AI and infrastructure, we can often connect founders with talent they wouldn’t find through traditional channels. And when it comes to future rounds, we’re proactive about making introductions to the right investors - people who understand the space and can add long-term value.

We trust our founders. Our goal is to be the most helpful, thoughtful voice in the room when they need it, and to stay out of the way when they don’t.

Judy: Have you faced any unique challenges or advantages as an Asian investor in VC?

Jenny: Being Asian is a huge advantage in VC, especially in AI. 

If you look at the landscape, more than half of the people building in AI are Asian.  I’m Chinese American myself, and that shared cultural background creates an immediate sense of trust and relatability with many of the founders we meet.

Take Kepler AI, for example. We were introduced to the CEO through an Asian American LP who’s close friends with him. That kind of community connection is powerful and it’s not just about warm intros, it’s about shared values, mutual respect, and a deeper understanding of each other’s motivations.

Beyond AI, you see this trend in other frontier sectors too, like biotech, robotics, semiconductors. These industries are heavily influenced by Asian talent, and being part of that demographic gives us better access to founders, technical experts, and even early-stage deal flow that others might miss.

I also think there’s a generational shift happening. A decade ago, many Asian immigrants were focused on survival and stability: getting citizenship, securing a job, building a life. Entrepreneurship wasn’t always the first option. But now, we’re seeing second-generation and more affluent first-generation Asians stepping into founder roles. They’re more confident, more resourced, and more willing to take risks. That shift is creating a wave of Asian-led innovation, and as an investor, it’s exciting to be part of that.

Ultimately, I see my background as a strength. It helps me connect with founders, understand their journeys, and support them in a way that’s both culturally and strategically aligned.

Judy: What trends in AI are you most excited about?

Jenny: One of the biggest trends I’m excited about is the rise of vertical AI agents. We’re finally at a point where models are good enough to go deep into specific domains, whether it’s healthcare, bioinformatics, construction, or legal tech. These agents aren’t just general-purpose tools; they’re tailored to solve real problems in complex industries. That’s where I think the next wave of meaningful innovation is going to happen.

We’re already seeing major players like OpenAI and Anthropic double down on verticals. OpenAI is pushing hard into healthcare, and Anthropic is fine-tuning Claude for more specialized use cases. This year feels like the beginning of a broader shift - away from generalist AI tools and toward deeply integrated, domain-specific solutions.

I’m also watching the agentification of enterprise workflows. While adoption isn’t quite there yet, I think we’re two to three years away from seeing AI agents embedded across organizations, handling repetitive tasks, automating decision-making, and even managing internal systems. That’s going to change how companies operate, especially in sectors that rely heavily on junior or entry-level roles.

Of course, there’s a flip side to all this. As these agents get better, they’ll start replacing a lot of those entry-level jobs. It’s exciting from a tech perspective, but it also raises serious societal questions. What happens to the junior bioinformatician when an AI agent can do their job faster and cheaper? I think we’ll still need humans in the loop, but the scale and structure of teams will look very different.

Judy: How is AI changing your work as a VC?

Jenny: I spend 5-8 hours a day using tools like ChatGPT, Claude, and Cursor. I don’t even use traditional search engines or productivity tools anymore. AI has become my default interface for writing, researching, and decision-making.

One of the biggest shifts has been in deal sourcing. We built an internal AI-powered platform that surfaces interesting companies based on our thesis areas. Instead of manually combing through databases or relying on inbound pitches, we can scan hundreds of companies in minutes. When I bring on interns, I don’t ask them to go out and find deals; I ask them to use our system, review the AI-curated pipeline, and flag what’s worth a deeper look.

Diligence has also become dramatically faster. If I want to understand a technical concept, benchmark a company, or draft a first-pass investment memo, I can do that in minutes with AI. I can ask nuanced questions, get structured answers, and iterate quickly. 

Even portfolio support is evolving. We use AI to help founders with everything from market research to product strategy. And internally, we’re experimenting with agent-based systems to manage workflows, track portfolio metrics, and even simulate fundraising scenarios.

But it’s not just about efficiency. It’s about amplifying judgment. AI helps us think better, faster, and more creatively. And in a world where everyone has access to the same data, the edge comes from how you interpret it and how quickly you can act on it.

Judy: What’s one piece of advice you wish you had when starting in venture?

Jenny: I wish someone had told me early on how important it is to build a personal brand and differentiate yourself. Venture is a crowded space.  If you don’t have a clear point of view and a way to stand out, founders won’t remember you, and LPs won’t know why they should back you.

For us at Leonis, publishing technical research has been a game-changer. Founders often tell us they’ve read our pieces before we even reach out. That kind of inbound credibility is hard to build, but it compounds over time. And it’s not just about marketing, it’s about being part of the conversation and shaping how people think about emerging technologies.

My advice would be: Start early in building your brand. Be intentional. Figure out what you want to be known for, and make sure your work reflects that. 

Judy:  If you weren’t in venture, what do you think you’d be doing today? Any unexpected career paths you’ve considered?

Jenny: I’d probably still be at OpenAI or pursuing academia. I actually just finished my PhD this past May, which made my parents very proud, especially after I originally dropped out to join OpenAI. My advisor even encouraged me to go on the academic job market, and if I weren’t in VC, I think I’d seriously consider becoming a professor. Research has always been a big part of my identity, and I love the idea of contributing to the field in a more formal academic setting.

I’m also a huge fan of strategy games and I’ve been designing games since I was a kid. In middle school, I used to create board games and narrative games for my classmates. I’d keep score in a notebook and run the whole thing like a mini studio. If I ever took a totally different path, I could see myself building games, especially ones that involve strategy, systems, and storytelling.

Judy: Final question - if you could have dinner with any AI pioneer, who would it be and why?

Jenny: I’d love to sit down with someone from the team behind the Colossus supercomputer cluster. What they’re building is absolutely fascinating from a technical perspective. The scale of compute they’ve managed to assemble, the architecture, the networking - there’s so much innovation happening under the hood that most people don’t even realize.

As someone who’s a bit of a GPU nerd, I have so many questions. How are they connecting and optimizing thousands of GPUs to train massive models? What kind of infrastructure challenges are they facing? How do they think about scaling, redundancy, and energy efficiency at that level? It’s not just about building powerful models, it’s about building the systems that make those models possible.

I think the Colossus team is pushing the boundaries of what’s possible in AI infrastructure. They’re not just advancing model capabilities, they’re redefining how we think about compute itself. That kind of work is foundational, and I’d love to hear their perspective on where things are headed, what bottlenecks they’re solving, and how they see the future of large-scale AI training evolving.