Company profile
Reward AI
Reward AI builds general-purpose robot intelligence that learns manipulation directly from natural human demonstrations. Its Omnibody stack combines a wear...
Facts
- Segment
- ROBOT_LEARNING
- Country
- United States
- Headquarters
- Palo Alto, California
- Website
- https://rewardai.com/
About Reward AI
Reward AI develops robot-learning systems around a common interface between human demonstrations and different robot bodies. Its Omnibody Hand is a wearable seven-degree-of-freedom device that records visual, tactile, proximity, pose and force information while people perform manipulation naturally. One Data Interface turns those signals into a consistent training format without requiring operators to teleoperate a particular robot.
OM-1 is the decision-making core of the stack. The policy learns directly from this human data and produces motion, speed, force and grasp timing for industrial arms, wheeled mobile manipulators and legged humanoids. A reinforcement-learned control layer runs beneath the policy at high frequency, compensating for dynamics, delays and disturbances while smoothing successive model outputs. Reward AI is also extending the shared-model approach toward coordinated multi-robot manipulation.
Summary
Type: Robot AI company
Core: Omnibody stack
Function: Learn cross-body skills
Audience: Robot operators