Builds software infrastructure for physical AI. It models chips as digital twins and uses AI agents to automatically generate the integration code — board support packages, device trees, and drivers — that lets software run on hardware, collapsing weeks of manual engineering. Headquartered in London.