Veeda AI Raises Over $90 Million to Train Robots in Simulated Worlds

Founded by Sanja Fidler, former head of laboratory at Nvidia, the Toronto-based start-up has signed one of the largest seed rounds ever completed in Canada.

Veeda AI is emerging from the shadows with more than US$90 million in seed funding, led by Khosla Ventures and Radical Ventures. The company was incorporated less than three months ago by Sanja Fidler, who headed Nvidia's Space Intelligence Lab, along with two former colleagues, Huan Ling and Zan Gojcic. The amount puts the deal beyond the 99th percentile of seed rounds in the industry and makes it one of the largest in Canadian history.

The project focuses on "world models": multimodal foundation models trained on data from sensors and physical scenes, capable of simulating the behavior of reality. The goal is to create virtual environments in which embodied agents — industrial robots, mobile machines — learn and are evaluated through interaction, before any deployment.

The economic argument is that of data: learning on a real robot remains slow, expensive and sometimes dangerous, which limits the volume of experience accessible. Moving part of this learning to simulation means lifting this ceiling — provided that the simulator is faithful enough that what is learned there can once again be transposed to a real machine.