I'm at the Jackson Laboratory in Farmington, Connecticut, working on two problems: cell fate decisions and mouse behavior. On the cell side, I care about the geometry of how cells commit to a fate — the trajectories, the basins, what happens when you perturb the system. On the behavior side, it's closer to forward genetics: phenotype mice, find structure in what they do, trace back to the genome.

Before this I was a mechatronics engineer, then spent a few years in industry teaching myself physics at night before a PhD at IISER Mohali on dynamical systems and chaos. My first postdoc was with Ulrike Feudel at the Institute for Chemistry and Biology of the Marine Environment in Oldenburg — a small research town in northwest Germany, the North Sea not far, grey skies most of the year. I worked on synchronization in food webs there. Toward the end of that postdoc I was already looking for problems at the intersection of machine learning and dynamical systems — it felt like the natural next direction. Around that time I heard from Bill Ditto at NC State; he was looking for someone to help shape his lab toward AI × physics, which sounded like exactly what I was after. I joined and spent three years there doing that.

After working through a range of problems at that intersection, I wanted something with different stakes — real messy data, and work that actually matters to people outside the field. That's what led to JAX.

Predrag Cvitanović says a scientist owns the method, not the subject. That's the closest thing to a plan I ever had.

What I'm up to now is on the research page. (I also drink an indefensible amount of tea—the one variable I've never managed to control.)