Companies that started
as research

Protein design models that help industrial biotech teams engineer enzymes in fewer wet-lab cycles.

Generative models that propose and rank candidate battery materials before synthesis.

A compiler that auto-tunes ML kernels across GPUs and custom accelerators, without rewrites.

Evaluation and regression testing for models in production, so teams know when a new model is actually better.

Weather and flood nowcasting models built for the tropics, sold to insurers and city operators.

An inference runtime that uses sparsity to run large models on edge hardware.

Orchestration software that lets wet labs run experiments autonomously, guided by active learning.

Retrieval and data infrastructure for scientific literature, figures and lab notebooks.

Differentiable optics simulation for photonic chip design.