17 August 2026
Abstract: Autonomous vehicles are already demonstrably safer than human drivers yet no regulator can certify this claim, in the same manner that we certify elevators or autopilots, because a learned driver’s behavior is demonstrated statistically, not specified in advance. The same gap opens wherever Physical AI gets embodied – in warehouses, on factory floors, in the home – the moment a robot’s behavior comes from an opaque model rather than a specification.
This talk presents Continuous Statistical Assurance (CSA): a framework where the certificate itself becomes a live, machine-readable safety envelope that providers and regulators read identically, in real time. Ref. Paper: Certifying the Black Box: Continuous Statistical Assurance for Autonomous Vehicles — DOI: 10.5281/zenodo.21438788
Biography: Neel Shah is a product and systems leader working at the intersection of AI and safety-critical hardware. Over fifteen years at Jaguar Land Rover and Lucid Motors, he has taken vehicle programs from concept to production – including the Jaguar I-PACE platform now deployed as Waymo’s autonomous fleet vehicle – and built early brake-by-wire and steer-by-wire (ISO 26262 ASIL-D) systems.
At Lucid, he led technical product planning for a greenfield >3B vehicle platform and a 150-person organization, and shipped the company’s first production AI workflow. He is the author of Certifying the Black Box: Continuous Statistical Assurance for Autonomous Vehicles (2026). Neel is a Chartered Engineer (IMechE, UK), now focused on bringing trust to Physical AI and robotics. This session will be recorded.