PERCEIVE
Transform multimodal sensor streams into structured representations of people, objects, and space.
imagine.
Intelligence touching the world.We turn the unseen into understanding.
And understanding into action.
We develop world models that understand complex, real-world environments and enable intelligent systems to take useful, generalizable actions.
Transform multimodal sensor streams into structured representations of people, objects, and space.
Learn how environments evolve, how actions change them, and what outcomes are likely to follow.
Convert intent into robust plans and embodied behaviors that transfer across tasks and scenes.
Models that reason over dynamic environments and connect perception, prediction, and action.
→02Generalizable robot learning for manipulation, navigation, and physical interaction.
→03Unified learning across vision, language, motion, and diverse real-world sensor data.
→Join a team working across research, infrastructure, hardware, robotics, and product. Choose the straightforward careers site—or explore roles through our interactive Talent System.