AI researcher Sanja Fidler raises US$90-million for robotics startup

Former Nvidia Corp. NVDA-Q researcher Sanja Fidler has launched a new Toronto startup that will build artificial intelligence models to help train robots, with backing from venture capital firms Khosla Ventures and Radical Ventures.

Ms. Fidler is chief executive officer of the new company, which is called Veeda Innovation Inc. and will have offices in Toronto, Zurich, Singapore and Mountain View, Calif. She co-founded the company with former Nvidia colleagues Zan Gojcic and Huan Ling.

“We started Veeda because we believe robotics will reshape the world,” she wrote in a LinkedIn post on Wednesday. The company was incorporated in June, a month before she announced her departure from Nvidia, where she most recently worked as vice-president of AI research in Toronto.

The company has raised more than US$90-million, according to a source familiar with the matter. The Globe and Mail is not identifying the source because they are not authorized to speak publicly on the matter.

A spokesperson for Veeda declined to comment.

Advances in AI have ignited the robotics sector in recent years. Companies such as Tesla Inc. TSLA-Q, Figure AI and 1X Technologies are building general-purpose humanoid robots to someday work in factories and perform domestic chores inside the home. Chinese company Unitree Robotics went public in Shanghai on Wednesday, with shares rising by around 460 per cent. Some auto manufacturers are already testing humanoid robots inside of assembly plants.

These robots require sophisticated AI models in order to interact with the world and perform tasks autonomously, rather than be teleoperated by a human.

Veeda will focus on building world models that mimic physical reality. One of the challenges in the field is that training a robot in the real world can be expensive, slow and carry safety risks. Veeda will instead create a simulated training environment “where robot intelligence can interact with an entirely virtual world in a million possible ways, learning from its own mistakes,” according to its website.

The company referred to its simulator as The Matrix for physical AI, referring to the science fiction film.

Toronto-based autonomous vehicle company Waabi Innovation Inc. has also trained AI models in simulated environments, which founder Raquel Urtasun has argued is less expensive and safer than on-road testing.

Nvidia has become a US$5-trillion company because of demand for its chips, but it has long conducted AI research, and Ms. Fidler has played an important role. Considered a star in Canadian AI, she grew up in Slovenia, where her father inspired her with stories of famed inventor Nikola Tesla. “This appealed more than a movie celebrity or a singer,” she told The Globe and Mail last year.

She completed a postdoctoral fellowship at the University of Toronto in 2012 and then became an associate professor before she was tapped to run Nvidia’s Toronto research lab in 2018.

Much of her work is connected to world models, spatial intelligence and simulated realities. That includes GameGAN, a generative AI model released in 2020 that can reproduce a video game based on screenplay and keystrokes. The following year, she helped develop DriveGAN, an AI-powered driving simulator based on 160 hours of dashcam footage.

She was also part of the team that developed Cosmos, a platform for building customized world models for robotics and autonomous driving. First announced in 2025, the platform has been used by Uber, Magna and Figure AI, according to Nvidia.