Datoric

Datoric develops training data and custom collection services for robotics and physical AI, aiming to build a foundational data layer for systems that operate in the physical world. The company offers licensed, ethically sourced datasets with documented provenance and quality review, as well as tailored data collection for specific research or model-development needs.

Its data modalities include egocentric video of household and workplace tasks labeled by task, action, and object; bimanual UMI demonstrations with left and right gripper views, end-effector poses, and gripper width; and teleoperation recordings synchronized across video, states, actions, robot platforms, and sensor configurations. Other listed datasets include annotated dual-camera manipulation, tactile-glove and stereo manipulation, and paired egocentric RGB and IMU data.

Datoric also investigates data-related causes of model failures and capability gaps. Its process uses controlled collection experiments across video, motion, and tactile modalities, followed by downstream evaluation to test whether interventions transfer. It then specifies validated methods for production. Backed by Y Combinator, Datoric invites partners to explore datasets, discuss research, and plan collections.