Roche

Enzyme function assignment by embedding retrieval, fusing three frozen protein language models into a 256-D EC-aware search space.

Chemistry embedding model placing SMILES strings and scientific prose in one vector space, top-tier on 43 of 48 molecular and text benchmarks.

Antibody, nanobody, and T-cell receptor structure prediction that resolves bound and unbound conformations separately in under a second per domain.

Patient-level foundation model that pools every cell in an scRNA-seq sample into one disease representation, trained on 24.3 million cells.