E
Cambridge, Massachusetts, US
PI
Physical Intelligence builds foundation models and learning algorithms enabling robots to adapt across tasks, environments and hardware with minimal custom programming.
1 open job
Eka
Eka develops robots with force-based intelligence for dexterous manipulation, combining vision, force and action models to master physics across food handling and manufacturing.
Eka builds intelligence for the physical world, treating force as a first-class representation of intelligence. The company develops robots that master physics and perform dexterous manipulation tasks, beginning with problems long considered unsolved in robotics. Its Vision-Force-Action (VFA) model combines generality and performance by representing forces directly, aiming to push robots beyond human limits in handling the physical world.
The work spans force-based physical intelligence, dexterous manipulation, self-supervised learning, large-scale robot learning, world models, reinforcement learning and sim-to-real robotics. Demonstrations include sorting chicken nuggets and screwing in light bulbs, with motion Wired described as "eerily lifelike". The approach targets robots that are fast, general and reliable, rather than narrow automation.
Applications range from food handling and processing to manufacturing, assembly and automation. The team is drawn from MIT, Berkeley, Harvard, CMU, BU and UPenn, with prior experience at DeepMind, Microsoft and Boston Dynamics.