Hardware design in GDPx
Our contribution to SemiAnalysis' GDPx benchmark, evaluating agents across CAD, drafting, CAM, and EDA.
A few beautiful machines have shaped the course of history. We build AI systems so engineers can create the next ones.
We imagine a future where engineers work more like architects: defining requirements, evaluating concepts, and owning difficult tradeoffs across form, fit, function, and manufacturability. Agents elaborate those decisions into detailed geometry, iterative simulations, and production-ready artifacts. We're reinventing CAD/CAE infrastructure from the ground up for this new division of labor.
Our team is making the journey from idea to manufactured product fully programmable. We’re starting with Tau, a CAD/CAE system built from the ground up for human–AI collaboration, where code and UX are isomorphic, simulation-based verifiers are baked in, and agents can explore designs with massive parallelism. Conveniently, our stack doubles as the most efficient reinforcement learning environment for engineering. And using it, we’re pushing the frontier of intelligence.
Read moreWe collaborate with frontier labs to develop reinforcement learning infrastructure and benchmarks for real engineering work. Our applied research is conducted alongside the enterprise teams closest to these problems, so progress is measured against what excellent engineers actually do.