Laplace Intelligence uses AI to stress-test robots before deployment

Founded by CEO Howie Sun and chief scientist Henry Liu, the startup emerged from University of Michigan research that previously earned a cover feature in Nature. The company utilizes a simulation platform built on Nvidia’s Cosmos, which reconstructs real-world locations—from Arizona highways to Paris intersections—at scale. By integrating live data from providers like TomTom, developers can manipulate variables such as traffic patterns to observe how autonomous agents react to complex, dynamic scenarios.

Unlike traditional methods that rely on human-designed edge cases or static footage, Laplace uses AI agents to autonomously generate rare, tricky scenarios based on police reports and historical collision data. This approach shifts testing from a manual process to a dynamic, interactive loop. The startup has already secured contracts with clients including Bot Auto and a major Japanese automotive manufacturer, while collaborating with T-Mobile and Nvidia on a smart city pilot in San Jose.

While established competitors like Applied Intuition hold significant market share, Sun argues that the shift toward AI-driven, interactive testing represents a new paradigm for the industry. As the team prepares to present at TechCrunch Disrupt, their focus remains on scaling these simulation capabilities to support the emerging field of generalized robotics, where real-world data remains scarce and edge cases are still being defined.