Tyvori’s Physical AI Data Layer

Field notes

Tyvori describes a contact-led acquisition workflow for robotics teams: define a physical skill, source a relevant environment, design the capture protocol, record demonstrations, and deliver structured data. The listed inputs extend beyond ordinary video to RGB-D, IMU, pose, trajectory, force, contact, grip, and tactile signals. The page also separates successful demonstrations from failure and recovery sequences and from evaluation datasets. These are presented as service capabilities rather than independently verified completed programs.