Founder Spotlight #15: Charles Fisher @ Unlearn
BIOS: Nucleus of Life Science Innovation 🚀
JOBS
- BIOS Talent: Find Jobs @ Breakout TechBio Startups — Search Jobs 🚀
- Post Jobs: Add Your Startup to BIOS Talent — Post Now 🎉
- Students: Join Alix Ventures Fellowship — Join Now 🧬
- BIOS Contributor: Share Your Thought Leadership — Join Now🔬
CONTENT & COMMUNITY
- BIOS Daily: Join 25K+ Subscribers Following TechBio — Sign Up 🔥
- BIOS Insider: Premium TechBio Thought Leadership — Sign Up ✨
- BIOS Commons: World’s Largest #TechBio Community — Join Now 🎉
INVEST
- BIOS Angels: 1st TechBio Angel Investing Syndicate — Join Now 🌟
- Alix Limited: Invest in Breakout TechBio Startups — Learn More 🧠
By:
Alix Ventures: Supporting Early Stage Life Science Startups Engineering Biology to Drive Radical Advances in Human Health
Calling All Innovators — Click to Reach Out 🚀
Charles Fisher is a scientist with interests at the intersection of physics, machine learning, and computational biology. He is currently Co-Founder & CEO @ Unlearn, an artificial intelligence company working to speed up clinical trials. Previously, Charles worked as a machine learning engineer at Leap Motion and a computational biologist at Pfizer. He was a Philippe Meyer Fellow in theoretical physics at École Normale Supérieure in Paris, France, and a postdoctoral scientist in biophysics at Boston University. Charles holds a Ph.D. in biophysics from Harvard University and a B.S. in biophysics from the University of Michigan.
Unlearn, based in San Francisco, CA, accelerates drug development by adding machine learning-generated digital twin data to the control arms of clinical trials. We generate digital twins for all patients enrolled in a clinical trial, regardless of randomization assignment (i.e., there are digital twins for both active and control patients). Digital twins are clinical predictions of the outcome if a patient was assigned to the control arm.
Personal Spark
What prompted you to pursue a career in Life Sciences? Was there a specific moment in time or influence you can remember? What drives you to work in this space?
My pathway to starting Unlearn is pretty winding. I started college at the University of Michigan undecided. The summer after my freshman year, I got a research job in Radiology at Michigan State. Working with MRIs during this experience helped me realize that I was a lot more interested in the physics behind medicine than the medicine itself, so I went on to study biophysics for undergrad and my PhD. Throughout this time, I was working at the intersection of developing machine-learning based methods and applying them to biology. Both pretty broad concepts, I worked on projects ranging from drug discovery and protein folding to ecosystem mapping for butterflies. After my post-doc, I took a job in computational biology at Pfizer. One of the projects I was working on involved applying machine learning to clinical trials to predict treatment responses. That got me interested in clinical trials, but I actually left biology for a short-while to work at LeapMotion, a virtual reality startup in San Francisco. Everything happens for a reason and even though I quickly realized I was more drawn to solving problems in biology, I wouldn’t have met my co-founders if it wasn’t for LeapMotion.
Company Overview
What problem is your company solving?
At our core, we’re a machine learning technology company. On the life sciences side, there are two main problems we’re addressing. In the short run, our focus is on making clinical trials faster. For example, a typical trial for Alzheimer’s Disease takes 4–5 years. Of that, only one year is actually spent following a patient. The other 3–4 years is spent just finding patients to fill the trial. Unlearn is focused on taking data from existing trials to make this process quicker and more insightful.
Our long-term goal is to create what we call Digital Twins — which is essentially a computer simulation of a patient. With a Digital Twin, we could answer clinical questions through simulations, figuring out insights such as what would happen to person X if they were given the standard of care. All our work with machine learning is laying the groundwork to create this type of technology.
How did you become motivated to solve it?
We were initially motivated by the machine learning aspect. I’m a theoretical physicist and my co-founders either studied theoretical physics as well or mathematics. At the time, almost all of machine learning was driven by a handful of companies in the tech industry. The problem with that was we were using machine learning to solve problems only in tech, leaving a whole field of applications that could implement ML. Understanding and learning from clinical data was one. That being said, we ended up approaching the problem in a reverse way. We knew we wanted to focus on the clinical data, but instead of applying a machine learning algorithm to clinical data, we realized that if we look at the clinical data itself, we could come up with interesting ML algorithms to make sense of it.
Now in technical language, what are the specifics of what your company does?
With Digital Twins, we are a generative modeling company. What we want to be able to do is generate hypothetical medical records for patients. This is particularly challenging because the data is longitudinal — we have multivariate data that deviates from the typical format of images. We currently use an unusual approach to ML by using principles from Deep Boltzmann Machines, which didn’t scale well for image data but actually works great for clinical data. We started off by using these machines to develop time series models and then combined them with newer methods to make a hybrid Boltz-GAN machine for time series. From a technical standpoint, Unlearn is unique in that we use a completely different type of neural network that is not found elsewhere.
Why does your solution matter for the world when you get it right?
We are working to reduce the number of patients required to run a successful clinical trial. Directly, this helps speed up clinical trials, of which we saw the need for this year with the pandemic. If we are successful in this, we would be able to speed up clinical research across the board and lower costs.
Genesis
How did everything come together and how did you meet your co-founder?
My co-founders, Jon and Aaron, and I were all working at LeapMotion before we ended up leaving the company at around the same time. We spent around four months throwing around ideas on starting a company that uses ML in areas where ML had not been implemented. We settled on clinical trials and have been running with it since.
Timing is everything — how did you know the timing was right?
This is a tricky question because if you knew the timing was perfect, everyone would already be doing it. We knew that there were certain areas of machine learning that were underserved. There was beginning to be a movement towards taking data from electronic medical records and deriving insights from it. The industry was starting to develop and there was enough interest in machine learning that we were convinced it was the right thing to tackle. As a founder, you walk a fine line between holding onto the core beliefs for starting a company and learning from your customers or market on what is important. We didn’t start the company knowing we were going to look at speeding up clinical trials. We started the company knowing we would apply machine learning to underserved clinical needs.
Accomplishments
What are some of the notable milestones your company has achieved thus far?
We have published several papers on our algorithms. Each paper is a milestone in and of itself because it stands for new technologies that we are discovering and implementing. Last year, we had a meeting with the FDA, which is less common for smaller companies. This provided us with support and feedback, really validating our approach of using data to make clinical trials more efficient.
What company achievements are you most proud of?
Hands down, I am most proud of the team that we’ve been able to build at Unlearn. Each person is exceptional in what they do and we’re only going to be getting better as success begets more success. We’re a small company with very few resources, so being able to build one of the best scientific machine learning teams in the world is incredibly fulfilling.
What are some of the biggest hurdles ahead? How do these create points of value inflection?
There are two big hurdles that come to mind. The first, communication, is something that we can control. We are a technical company and work at the intersection of three heavily technical fields: machine learning, statistics, and medicine. Neither of these three speak each other’s languages, so we have to be cognizant about how we are telling our story to ensure that each audience understands and appreciates what we’re doing. The other challenge is that sales cycles in healthcare are incredibly long. This is something that is getting a lot better as the market speeds up, matching business plans.
Pay It Forward
What are some of the must haves for an early stage HC/LS startup in your eyes?
All startups need cash, but they especially need good cash management practices. Especially at the early stages, we want to make sure we’re spending the capital we raise on things that further our business. Additionally, earning trust is absolutely essential for any startup. You want to seek out the right people to talk to, be transparent, and publish science, which I find a lot of tech companies veer away from. Especially in markets where people have been burned before, a strong foundation of trust is key.
Can you demystify the process of what it was like to raise VC funding? What were the highlights & low lights? Any advice or words of wisdom for future founders?
Each round has been different. It always helps to have a warm intro, but perhaps the most important component is having a compelling story. The first pitch is luck, especially if you are a first-time founder. You need to make sure you are finding the right partner and someone who is willing to take a chance with you. Later rounds are a different ballgame. For us, the second round was hard in that it took a lot longer than we had anticipated. You have to talk to a lot of funds, especially if the market is starting to take off. I don’t know if I have much advice other than to get lucky. And by getting lucky I mean really take the time to find the right people who will help you succeed. Venture investors shouldn’t just be providing capital. These are people you will work with for at least ten years, so you want to make sure they are real partners that you trust.
What advice for managing and hiring a great team can you share?
To attract the right people, you need a great mission and you need to be doing interesting science. Surprisingly, a lot of companies have only one of them. Star scientists and engineers are motivated by the science and engineering itself. They want to make sure they are working on interesting problems, so it becomes important for founders to empower them to build such technologies.
📣 If you enjoyed this post please clap 👏 & comment 💬 to let us know
❤️ Thanks Amee Kapadia for your help in putting this together :)
Alix Ventures, by way of BIOS Community, is providing this content for general information purposes only. Reference to any specific product or entity does not constitute an endorsement or recommendation by Alix Ventures, BIOS Community, or its affiliates. The views expressed by guests are their own and their appearance on the program does not imply an endorsement of them or any entity they represent. Views and opinions expressed by Alix Ventures employees are those of the employees and do not necessarily reflect the view of Alix Ventures, BIOS Community, affiliates, and content sponsors.
Join BIOS Community 🎉
Become a member, continue the conversation, connect with like-minded Life Science innovators, access exclusive resources, & invite-only events…
Apply to Join — Membership Application
For More Interesting Content 💭
🧬 Podcast — Stream Full Episodes
🧪 YouTube — Watch Videos
🩺 Twitter — Explore Feed
🦠 LinkedIn — Read Posts
