Raquel Urtasun

CEO & Founder — WaabiProfessor — University of Toronto

Profile

Raquel Urtasun is founder and CEO of Waabi, a Toronto-based autonomous driving company making a bet that the industry’s brute-force approach is wrong. Before Waabi she was Chief Scientist at Uber ATG — running autonomous vehicle research until Uber offloaded the unit in 2020. She’s still a full professor of computer science at the University of Toronto and one of the most cited researchers in computer vision and self-driving. “Waabi” means “she has vision” in Ojibwe. The name fits.

The Waabi thesis is straightforward and contrarian: you can’t scale self-driving by logging billions of miles on public roads. Instead, Waabi builds a single end-to-end AI driver and trains it mostly in Waabi World, a neural simulator that turns real sensor logs into closed-loop, physically accurate simulations. Test in simulation, validate on real trucks, iterate fast. It’s closer in spirit to how frontier labs train LLMs than to how Waymo or Tesla approach the problem. For developers watching the robotics space, Urtasun’s bet is that foundation-model techniques — generative world models, end-to-end differentiable stacks — will beat the hand-engineered, HD-map-heavy pipelines that dominated AV for a decade.

Waabi went commercial on trucks first, which is the rational order: highway driving is a more constrained problem than urban chaos, and freight has clear economics. In October 2025 Waabi unveiled an autonomous truck with Volvo, integrating the Waabi Driver into Volvo’s VNL Autonomous platform. In early 2026 Waabi raised $750 million (Khosla Ventures, G2) plus a $250M strategic investment from Uber, expanding into robotaxis with an exclusive commitment to deploy at least 25,000 vehicles on Uber’s platform. That’s a serious vote of confidence from the one investor who watched the previous version of this problem up close.

For someone learning AI: Urtasun is worth following because Waabi is one of the cleanest public examples of “physical AI” — applying generative models and world-model learning to a real-world robotics problem with billions of dollars on the line. The research her team publishes (UniSim, Copilot4D, GoRela) is the kind of work that shows where the frontier of embodied AI actually lives, away from the chatbot headlines.