Machani Robotics
Humanoid RoboticsTraining data behind physical AI.
Egocentric video annotation, grasp-action labelling, and safety-critical scene parsing for Machani Robotics’ CeCe desktop companion and RIA humanoid care robot.
Publication in progress
The full case study is moving through publication approval.
The Machani partnership covers first-person and third-person video annotation for humanoid robotics — the data layer behind CeCe’s desktop interaction model and RIA’s care-environment scene understanding. Metrics and published detail are in approval with Machani Group leadership. In the meantime, we’ll walk through the programme under mutual NDA on request.
What we can share without approval
- • First-person (egocentric) video annotation for humanoid training — the modality most gig platforms cannot deliver consistently.
- • Third-person scene parsing paired with egocentric for training-data augmentation.
- • Grasp-action and object-interaction labelling with action segmentation.
- • Safety-critical scene classification for care-environment deployment (RIA).
- • A dedicated pod that holds taxonomy continuity across multi-quarter training runs.