Safeworld, a new Carnegie Mellon-linked startup, is tackling a central problem in robotics: generative AI models are powerful but not predictable in the way traditional control algorithms are. Founded by Safe AI lab director Dr. Ding Zhao, startup veteran Kyle Wong, and machine learning engineer Simo Rachidi, the company aims to underwrite the risk of probabilistic robot systems and build the trust needed for real-world deployment.
The company's approach is to build digital simulations populated with realistic human models, then run thousands of scenarios where the robot's actual software encounters unexpected behavior—such as a person tripping, crouching, or appearing from a blind corner. Gritt Robotics, which builds AI brains for solar-installation robots, is already partnering with Safeworld. Gritt's CTO notes that safety for such systems cannot be proven with equations alone; it must be tested empirically.
Safeworld is still deciding whether to offer its product as a platform or a service, but the team is confident there is demand. As Zhao put it, anyone who wants to deploy a robot will need to pay for safety validation. The seed round was led by Shine Capital and a16z Speedrun, with additional backing from Box Group, Carnegie Mellon University Endowment, Innovation Endeavors, and SV Angel.