Moonlake AI

Moonlake AI builds simulation environments and world models for robot development. Its platform helps robotics teams create digital twins, generate task-specific training data and evaluate robot policies in simulation. Its 3D Agent creates assets with geometry, articulation and physical properties such as mass, friction and collision behavior. Moonlake delivers editable, layered USD assets with validation scripts, allowing teams to inspect simulated interactions and revise objects as training exposes new failure modes.

Co-founded by Fan-Yun Sun, its CEO, and Sharon Lee, Moonlake combines robotics, graphics and simulation research. Its environment-generation approach develops assets, scenes and simulator settings together: an object's collision geometry and materials must work with the simulator's numerical settings to reproduce relevant physical behavior. The aim is to make simulated policy evaluations informative about real robot performance, rather than relying on visual resemblance alone.

Moonlake also studies how learned models can complement classical simulation. Its Reverie research includes Reimagine, which adds realistic appearance while retaining simulator-controlled geometry and state, and Daydream, which uses simulated robot motion to constrain predictions of future observations and states. In 2025, Moonlake announced $28 million in seed funding from Threshold Ventures, AIX Ventures and NVentures, NVIDIA's venture capital arm.