How Schrödinger sped up molecular discovery by 4x with Alphaevolve

Schrödinger partnered with Google Cloud to deploy AlphaEvolve, an AI coding agent, to speed up molecular discovery by 4x. They targeted two critical algorithms in their machine-learned force fields (MLFF) training pipeline: neighbor list computation and Ewald summation. By incorporating AlphaEvolve, the system generated a batched implementation of the Ewald summation, outperforming existing custom kernels. Schrödinger used a rigorous evaluation framework to confirm the evolved code was both performant and scientifically accurate. The evolved code achieved a 4x speedup in molecular discovery.