Sanja Fidler Launches Veeda AI for World Models

Veeda describes its work as building "the next generation of multimodal foundation world models for physical AI," according to its website, as cited by The Logic. BetaKit describes world models as AI systems trained on physical, spatial, and movement data about an environment.

In a statement reported by BetaKit, Fidler wrote: "At Veeda, our sole mission is to build simulated reality for Physical AI. Our conviction is that this will become the critical infrastructure layer for all areas of robotics."

The Logic reports that Veeda is recruiting scientists and machine learning engineers to build technology spanning language, images, and video, with applications in robotics and embodied intelligence.

World models are an active research and product-development area for teams attempting to give robots and agents predictive representations of changing physical environments. In comparable efforts, the central technical questions include the quality and diversity of training data, the ability to model temporal and spatial dynamics, and whether simulated predictions transfer reliably to real hardware.

For ML practitioners, the company's stated multimodal scope makes evaluation methodology especially consequential: useful physical-AI systems generally need to be tested not only on visual or language benchmarks, but also on long-horizon control, robustness to distribution shift, and real-world safety constraints.

Fidler recently left Nvidia after eight years, according to The Logic. She is also a University of Toronto professor, and BetaKit reported that she had written earlier in August that world models are where the next breakthrough lies.