Laplace Intelligence, a University of Michigan spinout founded by CEO Howie Sun and chief scientist Henry Liu, is pioneering automated red-teaming for robotics and autonomous vehicles. Their goal is to stress-test physical AI systems in highly realistic, interactive simulations before they are deployed in the real world. This approach leverages AI agents designed to expose vulnerabilities and failure points in robots and self-driving cars by simulating real-world environments reconstructed from actual map data, including dynamic elements like traffic and pedestrian behavior.

The startup builds its simulations on Nvidia’s Cosmos platform, enhanced to accurately replicate real-world locations such as highways or city streets, using inputs like Google Street View and live traffic data from providers like TomTom. Unlike static or simplified models, Laplace’s simulations include multiple interactive participants that respond intelligently to each other, creating a dynamic environment to better predict how AI-powered machines will behave in complex situations. Their AI agents further enrich the process by autonomously designing challenging corner cases based on real-world incidents such as news reports or police collision data.

Laplace has successfully turned its academic research into a profitable business, attracting customers in the autonomous vehicle sector such as Japanese trucking manufacturer Isuzu, autonomous trucking firm Bot Auto, and technology giant Nvidia. The company is also involved in a smart city pilot in San Jose, California, and is exploring partnerships in humanoid robotics for warehouse automation. While many industry players have begun incorporating AI-generated edge cases for training, Laplace's fully AI-driven testing marks a new paradigm they believe offers a stronger competitive moat compared to rivals like Applied Intuition.

The startup is currently competing in the prestigious Startup Battlefield 200 at TechCrunch Disrupt 2026 in San Francisco, aiming to scale its operations and secure further funding. By integrating AI with physical AI testing in realistic, interactive simulations, Laplace is positioning itself to become a key player in ensuring the safety and reliability of robots and autonomous vehicles before they enter real-world environments. Their approach promises to build trust among customers, insurers, and regulators by providing solid evidence of tested robustness in physical AI systems.