Meet Lisha Li, CEO of Rosebud AI

Taylor Fang
Aug 18, 2020 · 15 min read

Key takeaways from Lisha’s journey across drama, pure math, investing and entrepreneurship

About

Welcome to the first installment of Tsingyuan Ventures’ Meet Our Portfolio portrait series. Every week, we publish an in-depth founder interview, ranging from early-stage entrepreneurs to successful businesses. Our conversations cover their personal journeys, the lessons that shaped them, their visions for the future, and their failures. We also learn more about their companies and the challenges they try to solve. These insights and lessons are applicable to any entrepreneur — current or future.

Rosebud AI

Lisha Li is the founder and CEO of Rosebud AI. Rosebud, founded in 2018, is a San Francisco-based synthetic media company. Rosebud creates images and videos programmatically with AI that look indistinguishable from ones made in a physical photoshoot.

Lisha was previously a principal at Amplify Partners leading investments in early stage startups, and worked in data science at Pinterest and Stitch Fix. Lisha holds a PhD in Machine Learning from UC Berkeley.

Why we invested in Rosebud AI: The commercialization of deep-learning based AI applications will accelerate in the next few years. We recently blogged about the “democratization of AI” as one of the tech shift trends we’re focusing on. Rosebud is a prime example. Lisha is a phenomenal entrepreneur with a diverse background across research, investment, and fashion. She’s well-poised to lead Rosebud to become the leading company in synthetic media.

Generative Photos demo via Rosebud AI YouTube

Meet Lisha Li

Interview edited for clarity and length.

“You have to build up mental stamina to be both extremely self-critical but also wildly optimistic.”

Lisha introduces herself

I’m Lisha Li, founder and CEO of Rosebud AI. It’s a little difficult to summarize my experiences, because it’s very hard to plan out a 10 or 20 year “life plan.” But one of the things that I solve for is what’s both intellectually interesting and personally challenging. I always try to stay true to that. As you change and if you surround yourself with different environments, you discover new things. You find that the things that are interesting and intellectually curious also evolve.

As you change and if you surround yourself with different environments, you discover new things. You find that the things that are interesting and intellectually curious also evolve.

I was very interested in fundamental science, especially physics and mathematics. It was an amazing human achievement to create these sources and bodies of knowledge, and also intrinsically deep and rich. It was interesting to me on many levels, but I also did a lot of artistic pursuits.

I chose to have very intense vertical experiences. I was very interested in drama and acting when I was younger and lived in Toronto. I thought, huh, what actually works there? I kind of figured out: how do you get into a movie or get into commercials? It was tinkering with how things work, sociologically. I figured out they have to have an agent. And so I got myself one and, and then was able to become a unionized actress who got sent out for various things. And this was all happening during high school.

I chose to have very intense vertical experiences.

These two more extreme and interesting experiences, and how I’ve tried to actualize them, are examples of how I like to pursue what is interesting to me without having too much of a 10-year plan on where they would lead.

On studying pure math

Mathematics really trains you. You have to think in a very fundamental way, and I felt that that was going to be applicable for a lot of other things. I feel like I’ve applied being able to exhaustively and comprehensively break down problems. That was a bit of foresight: for a lot of other things, maybe it’s easier to enrich on my own. But pure mathematics will open up a lot of other doors that I can evaluate later on. It would be really hard to amass that background if I didn’t pursue it formally.

Journey from math -> investment -> entrepreneurship

I came to Berkeley almost a decade ago for a pure math PhD. But halfway into it, I realized that the day to day of a math academic is very different from the pure intellectual interests that drew me into it in the first place. Having had more experiences in tech via internships, I realized that the cadence and the flavor of technical and product work were a lot more interesting to me. So I looked into working as a machine learning engineer.

But then I learned more about investing and venture and put myself in a position to be an investor at a venture firm, which was also a weird jump. But all of it was learning more about what types of careers would be a really good source for me to apply these skills I’ve accumulated, and also my interest in people and projects and tech. Entrepreneurship was the culmination of that.

[…] what types of careers would be a really good source for me to apply these skills I’ve accumulated, and also my interest in people and projects and tech. Entrepreneurship was the culmination of that.

On balancing exploration and planning

For a lot of these longer term goals, you have to plan in order to amass enough backdrops and context. But for me, it’s more of reassessing on a several year timeline to see if what I’m doing right now is both maximally satisfying and really allowing me to harness the creative interests that I have.

For me, it’s more of reassessing on a several year timeline to see if what I’m doing right now is both maximally satisfying and really allowing me to harness the creative interests that I have.

I do these check-ins fairly frequently. It’s how I end up pursuing a bunch of different opportunities, but also building a longer-term strategy around the question: what are the things that are important?

For example, a PhD program for pure mathematics is definitely not a month or even a year decision. It’s something that you have to spend your entire undergrad building into. Getting into that area was a lot more self-reflection, because when I started undergrad I was actually completely in history and and political science. But then I just did a little bit more self-reflection and discovery, as we all do in college, and then I was able to get a truer sense of what would be persistent for myself.

For example, a PhD program for pure mathematics is definitely not a month or even a year decision. It’s something that you have to spend your entire undergrad building into.

Generative Photos” via Rosebud AI website

Challenges of being a lone founder

Founding something is already hard. If you have somebody that you really trust and respect, you can share half the burden and motivate each other when things get tough. That would be the perfect scenario. The truth, though, is that finding the perfect co-founder is such a difficult task. As an investor, you see more examples before trying yourself. The best thing is having the perfect co-founder, but the thing that often happens is you have co-founder breakups that also break up the company.

Founding something is already hard. If you have somebody that you really trust and respect, you can share half the burden and motivate each other when things get tough.

In my case, I had the idea that I already was committed to pursuing. So I didn’t want to wait around for the perfect co-founder. I was open to finding co-founders within maybe the first 6 months of starting the company, because there was still a lot of room to share with the founding team. But by the time I reached six months into Rosebud, I already had a pretty solid product direction and people I was looking to hire for the team. It no longer made sense.

It’s hard. It’s really difficult. You bear a lot more responsibility for the people that you hire and for the vision. In order to be somebody who wants to improve, you’re self-critical as well. So you fall into those patterns. You have to build up mental stamina to be both extremely self-critical but also wildly optimistic.

You have to build up mental stamina to be both extremely self-critical but also wildly optimistic.

It sounds like a contradiction to hold. But I think that’s really what you need as a founder: to keep grinding it out and yet have a lot of self-discipline to not feel like being self-critical is going to make you depressed.

On culture and choosing people to work with

I always want to make sure I’m supporting a positive culture. It’s hard to make this fully codified as a smaller startup, but I do find that I pick out a couple of pretty concrete things.

1. I like people who are low in ego.

Obviously I love to attract smart candidates to the company. But people who are low in ego are very good listeners and willing to be critical in a constructive way. They’re okay giving their opinions constructively and being receptive to feedback.

That’s the best way for growth, not only on an individual level. In an early-stage startup, there are so many things that are chaotic that if you lose the ability to be transparent and to get feedback even from within, then you’re going to have even more trouble from outside. I think the low ego part is trying to encompass all of those characteristics.

If you lose the ability to be transparent and to get feedback even from within, then you’re going to have even more trouble from outside.

2. A lot of intellectual curiosity.

“Intellectual” might even be too narrow of a word. Although this is a technical startup, I do want that to also translate into wide-ranging curiosity. I think there is a general curiosity for how things work and how “cool” the tech is. But how does this translate into the right product? Do they ask a lot of questions and seek answers in an objective way?

I think there is a general curiosity for how things work and how “cool” the tech is. But how does this translate into the right product? Do they ask a lot of questions and seek answers in an objective way?

Intellectually curious people are always building and trying to learn in their outside extracurricular life. We have engineers who are learning different ways to market products. And so they search for random products to try to market. Curiosity shows that you’re always interested in growing yourself.

Those two traits encompass a lot of the cultural positivity I try to perpetuate.

Her best advice for founders

Talk to customers. You can always talk to more customers. As an investor, it’s one of those things you know and will say to people. But it’s really essential doing a startup because there are so many things that you have to get right. And especially if the tech is also quite hard, your iteration cycles could be much slower. But you have to resist that. You have to just iterate faster, and still talk to as many customers as you can. It actually can’t hurt under any circumstance.

Talk to customers. You can always talk to more customers.

People ask: why do companies pivot a lot? It’s like, no, this is what happens when you ask customers about what you built. You actually iterate. Sometimes it’s called pivoting.

The house is always burning until you get to product-market fit. And you just have to be brutally honest with yourself and with your assumptions. You keep on iterating with customers. I think that’s the true, honest story of being a founder. Things can look wildly different, but it’s just this tenacity to not falter or get demoralized and to have your eye on the prize.

The house is always burning until you get to product-market fit. And you just have to be brutally honest with yourself and with your assumptions. […] Things can look wildly different, but it’s just this tenacity to not falter or get demoralized and to have your eye on the prize.

How it Works”: Walkthrough of Rosebud AI app via Rosebud AI YouTube

On startup community

One of things I loved about Y Combinator is the fact that you’re sharing a cohort with so many other companies at your stage. And you also know other companies, who may be six months or a year ahead in terms of growth and fundraising.

These peers can be such a great support network because everyone is going through this craziness of being a startup founder, which is quite a rare experience. If you have a cohort of people, all having very similar problems dealing with the chaos of the day to day of running a startup, it’s a lot more of a morale-boosting community.

These peers can be such a great support network because everyone is going through this craziness of being a startup founder, which is quite a rare experience.

Being an investor-turned-entrepreneur

When I was investing, I became intimately familiar with what it means to build a venture scale business, to have customers build a product, ship things, raise money, and so on. That kind of work is very different from investing. But fortunately, by sitting on the investing side, you take a much more detailed peek into it. I just realized that building things was a lot more. That’s what I wanted to do rather than just investing. It’s very different from being a venture capitalist. I became an entrepreneur for this particular problem space that I’m in.

I became an entrepreneur for this particular problem space that I’m in.

On the intersection of creativity and AI

The Rosebud AI website describes Humans of AI (a project generating life stories for virtual beings) as a way to “explore our common humanity,” and calls it a collaborative art piece.

We’re very fixated on: these are all the bad things that can happen. Or how dystopian the tech is. But when you delve into it, I think most tech is neutral, it’s just how you choose to productize and how careful you are in choosing the right users for this in the right use cases that make it in a more positive or negative direction.

I don’t shy away from the problem if it has some not-obviously-positive use cases, because I feel like that means that more people should work on this to actually make something really useful and beneficial out of it.

[…] that means that more people should work on this to actually make something really useful and beneficial out of it.

There’s a lot of cool research on synthetic media and generating images. But getting it to work within the creative workflow, where you’re augmenting someone’s vision or you’re creating the ability to automate a photographer’s ability to generate a lot of images, that’s when you’re going to develop the right user base.

And so when I think about creativity, I think about themes of: how do you amplify somebody’s ability to do their work, but also, what is their goal in the beginning? They have some image in their mind and they want to create it.

How do you amplify somebody’s ability to do their work, but also, what is their goal in the beginning? […] Involving the human component is not just paying lip service to it. It’s necessary to creating a very successful product.

AI Editing App” via Rosebud AI website

Involving the human component is not just paying lip service to it. It’s necessary to creating a very successful product.

And then it’s also interesting from a storytelling and creative point of view. It’s like: wow, this tech is actually really good at surfacing many examples of some kind of domain. That’s something that humans are less good at. And so you can make these forces complimentary and it gets our product creativity going.

What makes Rosebud AI different

There’s a lot of research that makes cool demos, and it’s vastly developing this space of deep learning for synthetic media. But a lot of what would translate this technology into a product involves defining controls for the user. Even though we can produce a really realistic-looking face, how do you produce the face that somebody actually wants?

Even though we can produce a really realistic-looking face, how do you produce the face that somebody actually wants?

There’s a lot of stickiness and network effects in building out tools that are being used by people to define, for example, what is a good output? We’re really focused on making the technology actually usable, rather than quickly demo-ing.

I’m very fond of the Open Source research community and this arena, but I think that there’s actually quite a lot of work into productizing it so that creatives are getting the right kind of output that they need. So it means higher quality, and a degree of control of the actual output in terms of features that you have in your mind.

Her vision for Rosebud AI

In terms of a five year plan, I believe that this tech stack is going to make a lot of visual content creation wildly easier. For photographers or videographers, a lot of that stuff that they have to shoot in the real world would probably not be needed. You will actually be able to compose things that you can input into a computer. For example, make an image of a woman standing on a desert island with a certain backdrop. And you’d be able to produce that stuff not just for photos, but for videos as well.

Given that that will happen, what are the products and the order of operations to build products that will get us there? It would be an unfair advantage to actually collect the right kind of data and train on the right kind of workflows in order to produce something that’s very high quality. But also, what does that enable us to do once we’re there? I think this is something very positive for storytelling, and we want to make sure that it is used in these right ways.

Given that that will happen, what are the products and the order of operations to build products that will get us there?

Twitter

Twitter both makes me crazy and has actually been really useful. I’ve gotten investors on Twitter. Twitter is so useful for many things that LinkedIn is for. I built my Twitter audience on the deep learning folks. And so I consume most of my research knowledge through Twitter. There are other places you can go on. But Twitter is where people are amplifying a lot of the visual results in GANs. And so that’s actually a concrete use case. Mostly for the stuff in deep learning research, it continues to be high high signal to noise ratio.

Her podcast recommendation

The Portal by Eric Weinstein. I don’t think that’s new for a lot of people. But I’ve actually known Eric for over a decade back from the pure math and physics world. I’ve always appreciated his willingness to have deep conversations. He doesn’t shy away from controversy. And that’s a pretty good quality to have in this age.

Hobbies during quarantine

I don’t want to burn myself out because it’s very easy to keep doing all things Rosebud. It wouldn’t even necessarily burn me out because there’s so many creative activities that are interesting there. But in terms of true leisure activities, just last week, I booked a tech-enabled virtual personal trainer. I realized that I find it a little bit more demoralizing when I’m doing a workout if I don’t know it’s optimal for myself. I’m going to give that a try, three workouts a week that are supposedly personalized to me and which basically costs as much as just having a gym membership. And then playing some video games.

Her morning routine

I try to have breakfast with my husband before just starting everything. I think that’s really nice. And just taking that fifteen minutes to have your coffee and sometimes have pastries. Bedtime bleeds into daytime because everyone is working from home and they just get out of bed. You can just sit at your computer and start your day. I think just having some kind of way to segment is really healthy.

Bedtime bleeds into daytime because everyone is working from home and they just get out of bed. You can just sit at your computer and start your day. I think just having some kind of way to segment is really healthy.

One question she wishes was asked more often

It’s easy to fixate on the success stories. I would love to have more podcasts where it’s like: tell me about all the things you failed at. Just so people have a really good distribution, and so there’s not confirmation bias for success stories and confusion about which of the features actually matter. I would definitely tune in to that.

It’s easy to fixate on the success stories. I would love to have more podcasts where it’s like: tell me about all the things you failed at.

To get updates on this series, please follow our publication on Medium.

Follow us on LinkedIn: https://www.linkedin.com/company/tsingyuan-ventures/

Follow us on Twitter: @TsingyuanVC

Tsingyuan Ventures is a $100M seed-stage technology firm. We back technical founders across software, life sciences, and frontier technologies. Learn more about our origin story and our approach here.

Questions, thoughts, reflections? Let us know in the comments below. We’re always looking for great entrepreneurs and early stage ideas, and we’re always interested in having a discussion about venture, technology, and anything related. To see more about Tsingyuan Ventures, please visit our website: tsingyuan.ventures.

Foothill Ventures

Early-Stage Technology VC Fund Based in Los Altos, CA