Pre-seed investor. Recovering biophysicist. Writing on the epistemology of science and the ownership economy.
I'm co-founder and general partner at Cerulean Ventures, a pre-seed fund backing founders who apply AI to the physical economy: energy, supply chains, nature, agriculture, geospatial intelligence, industrial systems. We're often the first check, and sometimes the only one. I also built Scout, the agentic system that runs our firm, from its research engines over the scientific literature to the relationship graph it reasons with. Portfolio follow-on rounds have been led by Union Square Ventures, Index Ventures, Javelin, and Valor Equity Partners.
I trained as a physical scientist: two degrees at UC Irvine in chemistry and in biochemistry and molecular biology, then PhD studies in chemical physics at Illinois, where I built single-molecule spectroscopy instruments to watch DNA polymerases work, one molecule at a time. I left the lab, played poker professionally for three years through the financial crisis, and then spent a decade putting machine learning into products: Bayesian modeling over the medical literature at MetaMed, ML-driven ad serving at roughly two billion ads a day at Vungle (now Liftoff, NASDAQ IPO 2026), clinical AI reaching eight million patients at Conversa Health (acquired by American Well), and the healthcare marketplace at ZOOM+Care. In 2018 I co-founded Magalix, applying machine learning to Kubernetes operations and security; we grew it from zero to 25,000 developers and it was acquired by Weaveworks in 2021. Along the way I led the data science and ML research team at Typeform.
The through-line is watching how knowledge actually gets made, and what it costs when the process rots. The MetaMed work convinced me that much of the published medical literature doesn't survive Bayesian scrutiny, which became a decade of essays on the epistemology of medicine and science, work in the ownership economy on who gets to own the systems they depend on, and eventually the fund thesis: the most important scientific and industrial problems are becoming tractable to AI, and the people building there deserve backers fluent in both the science and the market structure.
Barcelona, mostly. Investing at pre-seed through Cerulean, and spending most of my research time on AI for science: agentic research systems, scientific-literature synthesis, domain-specific evals, and what it takes to make discovery loops run faster than publication cycles.
The podcast I co-founded and host on shared ownership, governance, and the economic models that distribute upside to the people who create it. 89+ episodes with founders, funders, and researchers, plus an annual summit of ~150 delegates.