Program · SCI·05
We build simulation and search systems across physics, chemistry, and materials, mapping spaces far too vast for experiment alone. Where a domain has laws, we teach models to respect them; where it has data, we let the models find what the laws imply.
What we work on
A program is a small set of hard, coupled problems. These are the ones this group is working on now.
First-principles simulation is exact and impossibly slow. Learned surrogates trade a controlled amount of accuracy for orders of magnitude in speed, fast enough to search, accurate enough to trust, and constrained never to violate the conservation laws we already know.
Most of chemistry and materials science has never been made. We treat discovery as search over a structured space, guided by models that predict properties before synthesis and by uncertainty that tells us where to look next.
The tightest discovery loop is one where the model proposes, the lab executes, and the result trains the next proposal, with as little human latency as the science allows.
Signals from the program
Selected writing from this program
Collaborate
We hire researchers and engineers who want to push one of these programs forward, and partners who want to put the results to work.
