According to reports, Laplace Intelligence, a spin-off from the University of Michigan, is leveraging interactive simulation and AI agents to conduct automated red team testing for physical AI systems such as robots and autonomous vehicles, proactively identifying scenarios that could lead to system failure before deployment. Founded a year ago by Howie Sun and Henry Liu, the company has already achieved profitability and counts clients including Isuzu, Bot Auto, and NVIDIA.

Laplace’s simulation platform is built upon NVIDIA’s post-training version of Cosmos, enabling reconstruction of road environments using real-world maps and simulating interactions among traffic participants; AI agents are designed to generate edge cases based on news and accident reports. The company is also collaborating with NVIDIA and T-Mobile on a smart city pilot in San Jose and is in early-stage partnerships with warehouse humanoid robotics firms. Laplace is currently seeking funding and scaling up, with its founding team set to participate in TechCrunch Disrupt’s Startup Battlefield.