New York-based startup Antioch has closed an $8.5 million seed round at a $60 million valuation to build simulation tooling for robotics developers. The platform creates digital twins of hardware complete with simulated sensors so teams can test edge cases and train reinforcement learning models without extensive physical testing.
The core technical challenge is sim-to-real transfer: making virtual physics match the real world closely enough that models trained in simulation actually work when deployed on hardware. Antioch draws on physics models from Nvidia and World Labs, adding domain-specific libraries on top. Current focus areas include sensor and perception systems for automated vehicles, farm equipment, construction machinery, and drones.
Antioch’s thesis is that within two to three years, autonomous system development will shift primarily to software, provided the sim-to-real gap can be closed.
