As a physics undergrad at Caltech, Eric Kelsic was drawn to the world of biology — to the idea of cells as machines that are constantly sensing, computing, and making decisions. With that, came the idea: what if we could understand biology enough to redesign new machines, new biomolecules with an outsized impact on human health?
Now, Eric is cofounder and CEO of Dyno Therapeutics, where they’re applying machine learning to engineer better AAV capsids, and ultimately, unlock the potential of gene therapies by solving in vivo delivery.
Eric joined me for a behind-the-scenes look at the origins of Dyno: from Eric’s start in academia, to the rationale behind a horizontal business strategy, leading with conviction, and disrupting the future of gene therapy. — Sajith Wickramasekara
* Editor’s note: The conversation has been minimally edited for length and clarity.
Building things and pursuing big (sometimes weird) ideas
Sajith Wickramasekara: Let’s start with you personally. How did you make the switch from physics to biology?
Eric Kelsic: My love for physics began because I wanted to pursue how things work. In undergrad, I found that I enjoyed research because of the mystery and problem solving. But the problems in physics felt very distant: far away literally, very small, very cold, or at some other extreme. It’s very different from the day-to-day world that we live in.
In high school, I really disliked biology, because I thought it was all memorization. Later on, I took a few courses that made me see it in a new way. What I didn't realize at first was that most of biology is a mystery we don't fully understand yet. Trying to understand how it works is fascinating to me.
I was also fascinated by the idea of building things, but felt that we didn't understand enough exactly how cells actually work. I wanted to learn how nature works and then based upon that, find ways to build new machines or molecules — even if I didn't know exactly how at the time.
Sajith: I feel like every entrepreneur’s got some kind of formative story where they were selling candy in elementary school or selling Pokemon cards. What was your version?
Eric: I have those entrepreneurial stories from childhood too. I sold arts and crafts, door-to-door. I started an origami business in my 4th grade class. But puzzle boxes really became an obsession of mine.
Puzzle boxes are usually made of wood, and there’s a secret to opening it. There could be sliding panels, internal pins, magnets — and so it becomes a sort of a physics problem, a research problem, and also an art. I started buying these on eBay, got to understand how they were put together, and thought, maybe I could make them more cheaply myself. That’s how it started.
Sajith: When did you know you wanted to start a company? Was it a dream of yours during grad school?
Eric: A little bit. I consulted for a few startups during grad school and Iearned that I really liked working on teams and solving hard problems with real impact.
The team-based aspect was very different from what I had experienced. In academia, you need to have your own project, or one that you're splitting with a co-author, and you need to keep things small so you can get the credit.
With a startup, the problem is so important that everyone makes a significant contribution, and the teams can be much bigger. The problems are hard, but you're solving them together.
Sajith: What about as a postdoc? The Church Lab seems to produce a high number of entrepreneurs per capita. What was unique about the experience, about having George Church as a mentor?
Eric: George is very encouraging — he’s a connector. He attracts people who come and want to do something big.
Sajith: What about giving feedback on which ideas have legs to go beyond an academic lab? Because a lot of times, people overfitting their PhD research to a company doesn’t always work out.
Eric: George is really great at pushing people to think bigger, or to take a big idea and break it down, and bring it down to earth. No matter how crazy the idea is, he’s willing to engage.
Sajith: That’s really important, because I feel a lot of great ideas start off looking wrong in some way, but if you rule it out too early, you won’t get anywhere.
Eric: Absolutely. In some ways, the best, most groundbreaking ideas are kind of weird in the beginning. But that’s how you find the key insight that’s going to unlock something new. George is very practiced in that, and I learned a lot about how to think that way from him.
Lessons from academia — and when to trust your own judgment
Sajith: What other skills from academia translated particularly well to building and running a biotech company?
Eric: Transitioning from academia to biotech, there was so much that I needed to learn. But being a beginner can be nice because you don't have the same objections that an expert might have.
Overall, there's a lot that transfers from doing research to building a biotech company. If you're trying to do something that no one else has done before, the scientific method is great for that. You have to be willing to challenge your beliefs and look at the evidence and change your mind. That’s something the best academics can do really well.
Sajith: What’s one of the biggest lessons you’ve learned?
Eric: I always tell PhD students, never believe that your adviser is more of an expert than you at your thesis project. Trust your own judgement and be willing to push back.
It’s something that also applies to starting a company. When you’ve been thinking about something for many years, you're probably a top expert in that area. Other people are going to tell you you're wrong, but they haven't been thinking about it so long and so deeply.
There might be many things that grow from that initial seed of an idea — if you have the conviction to pursue them more fully. And that's not just with the science or technology. It's also with regard to how the company operates or the culture of the company.
I definitely got a lot of pushback early on starting Dyno, since I had never worked at a company before. For example, a lot of things I’d do in a certain way that made sense to me were not the way things were commonly done. And about half of those were right and half of those were wrong. Figuring which half was the hard part.
“You have to believe that some of your ideas will be right — you just have to explore and find those opportunities. You can’t be afraid of doing things. But then you also have to be honest and recognize when something's not working.”
Building the prototype for Dyno — before AI/ML was cool
Sajith: Let’s talk about the founding insights for Dyno. What’s the mission?
Eric: We’re building high-performance genetic technologies that transform patient lives — and that starts with solving in vivo delivery. Today, we focus on engineering AAV capsids to optimize them for in vivo delivery. For example, to deliver to every cell in the brain, or muscle, or eye, in a very specific, targeted manner. Over time, with more efficiency, we can bring the cost of delivering therapeutic DNA down to zero.
Sajith: When you first started applying machine learning to engineering AAV capsids, was anyone else using ML like this? Or did people think it was crazy?
Eric: So we started thinking about this in 2015. Back then, there was very little machine learning being applied to sequence design. But I saw that there was definitely an opportunity coming.
Sequence design was the initial motivation for me. For my PhD, I did an experiment in a week, turned it into data, and then I spent two years analyzing that dataset. I knew that there was value in the data because you could just see the patterns literally. You could recognize them by eye — even though proteins are complex and, at the time, folding was thought to be impossible to solve.
And if you could recognize patterns from the data, then you could obviously make predictions. And so I thought, there's a data-driven wave of protein engineering that's coming — and I want to be a part of that, applying it to solve impactful problems in human health.
So what do we need to do for that? We can’t take two years to analyze every dataset. High-throughput screening isn’t enough; we also need high-throughput analysis. At that time, deep neural networks were becoming more commonly understood to be very good at automatically learning representations of data, in both images and audio. So I thought, why not proteins as well?
Sajith: What was the tipping point? When did it feel like this was going to be huge for therapeutics?
Eric: It was never a sure thing. It always felt like this was possible, and we should keep trying different things. I had worked on a lot of projects in my PhD and gone through the process of giving up on something when I felt it wasn't going to work.
Sajith: That’s a really hard thing to do. Products are often full of failure and knowing when to quit is a very difficult skill.
Eric: It is difficult. Sometimes it feels like innovation is right around the corner, so you don't know what you're giving up on — but there's an opportunity cost of working on the wrong thing. And mainly, what I learned was, the problem of working on too many things.
I really wanted to focus on one important problem and then to really push it to see where it could go. And I felt like maybe, capsid engineering and gene delivery was an important one.
So we had this new technology for protein engineering. Why not apply it there? That way we can make progress on this important problem, but also learn how to apply these technologies.
“By focusing on solving a hard problem, we can also learn to level up our technology.”
Platform over pipeline
Sajith: Bringing it to your business model, conventional biotech wisdom is to own your own gene therapies, and have your own pipeline. Instead, you make capsids and you partner with pharma. In the early days, was there a lot of pressure from investors to change that?
Eric: I definitely got those questions. Coming back to the need for focus, I wanted to know that I was working on the right thing. The worst case would be to spend all this time solving a problem, and then no one cares about it. So I did a lot of outreach from the beginning before even doing the experiments.
Sajith: Sometimes you see a tendency from entrepreneurs, especially in biotech, to want to keep things a secret and not talk too much about what they're working on. It sounds like you were very open about the problems you're trying to solve and your approach pretty early on with a lot of people.
Eric: I was pretty sure that if solving gene delivery with machine learning was worth doing, and people wanted to work on it, that it'd be faster to work on it together than to try to do it independently.
Sajith: Were you ever worried that you might communicate enough with someone for them to compete with you?
Eric: If that were true, then maybe it wouldn’t be the right problem for me to solve, right? The act of getting out there and talking about the idea was a good way of knowing whether this is something I’d have a good chance at succeeding at, versus someone else taking it and polishing it faster than me.
Sajith: So when investors told you, you should make a drug candidate, what was their reason?
Eric: It’s not that it was a bad idea. It’s just that they had never done things that way. They were used to doing things a certain way, and their initial reaction was, why don’t you think about it the way I’m used to?
That reaction was signaling to me that I hadn’t explained the opportunity clearly enough — why this would be worth investing in to do it differently.
Sajith: I think that's such an important point. With early-stage companies, often the first articulation of your idea is never the best one, and you have to evolve it over time.
Eric: I had to learn that. Over time, I had more conviction that the platform model was the right strategy.
Capsids are just one part of the gene therapy. The payload is going to be different every time because you're delivering a different therapeutic, but the capsid can be reused. You might not know what to expect when it's first delivered to a patient, but each time after, there’s substantially lower risk.
That's unlike many other platforms in biotech where you're testing new molecules, where every molecule has its own risks and the only way to know is with a human trial. It's the modular nature of gene therapy that enables this platform approach.
Sajith: So in theory, the more partners you have, the more the next partner benefits from your experience.
Eric: That’s right. We’re a development platform. We don’t want to compete with our partners; we want to provide them with a competitive advantage. This model has potential to be much more impactful, enabling us to stay focused on advancing the entire industry in these key technological areas.
“I would be happy if we could do for gene therapy what NVIDIA has done for AI. That is, pushing the state-of-the-art in a way that people are inspired to create new types of medicine or solve new diseases because they’re expecting the supporting technology to improve.”
Curating company culture
Sajith: I’m curious how you run the company. I’ve heard that Amazon has been an inspiration.
Eric: Because I’ve never worked in another company before, I always try to find ways to learn about how other companies work. Some of these practices we’ve made our own. One is an emphasis on written culture, which resonated with me coming from a scientific background.
For important decisions, a written narrative is the preferred way to present thinking, gather feedback, refine thoughts, and make a commitment to move forward. We started this during COVID, but even now it allows us to be more rigorous in our thinking and planning, at a larger scale.
The secret to successful partnerships and problem solving
Sajith: You’ve signed some significant partnerships with large, successful pharma companies like Roche, Novartis, Astellas. What lessons have you learned?
Eric: I was very much inspired by the Paul Graham motto, “Make something people want.” That was the start of all our partnerships. They’re committed to gene therapy, and we believe in its potential.
In many cases, partners had gene payloads ready to go, that they’ve been working on for years. But delivery remained a challenge. We, at Dyno, are focused on solving that challenge.
Sajith: I think a lot of entrepreneurs would love to know, what goes on behind the scenes of these big deals?
Eric: From the very beginning, we set up joint research committees. It’s about partnership success, and leading with the principle of “same team, same goal.”
That’s been key because we’re doing science — not everything’s going to work for the first time, and we’re going to have to solve problems together. When we see a problem, we don’t hide it. We say, here’s what we see as solutions, and here’s what we’re committing to do for the future. We want to invite our partner into that process to find solutions together.
Advice for aspiring entrepreneurs
Sajith: Any advice for scientists aspiring to start a company? What would you tell your younger self?
Eric: There's no wrong way to start a company. So if you have conviction, then don't let anyone tell you you’re wrong. Just have a mindset where you're able to learn and correct quickly.
What's worked for me is thinking about what I was uniquely qualified to do, or even uniquely motivated to do. Also, taking your time and being patient to find the right type of problem, and then verifying that it is the right problem.
