01
Review layer included
Demonstration data
Task trajectories, action segments, corrections, and natural-language instructions.
Embodied intelligence
Training signals for robots learning from the physical world.
Human demonstrations, perception data, and task evaluations for manipulation, mobility, and embodied agents.
Signal map
Program outcomes
Useful demonstrations
Environment diversity
Task-level performance evidence
Inside the work
Human expertise, connected to a controlled data workflow.
Every engagement links the people doing the work, the evidence used to review it, and the model behavior the program is meant to improve.
Capabilities
Built to fit the model, domain, and decisions behind your program.
02
Review layer included
Perception labels
Objects, affordances, states, contacts, and spatial relationships.
03
Review layer included
World-model data
Temporal sequences and outcome labels for prediction and planning.
04
Review layer included
Embodied evaluation
Repeatable tasks and rubrics for completion, efficiency, recovery, and safety.
What gets delivered
More than a dataset: a usable package of data, controls, and learning.
The exact artifacts change by program, but every delivery is designed to be inspectable, actionable, and ready for the next model decision.
01 / Design
Ontology package
Label definitions, examples, ambiguity rules, and a controlled change log.02 / Data
Validated annotations
Structured labels with review status and difficult-sample routing.03 / Insight
Coverage report
A clear view of edge cases, disagreement, and taxonomy performance.Delivery model
Designed for fast learning and controlled scale.
Programs move in visible stages, with a review point before scope, volume, or complexity increases.
- 01Frame
Decompose the task
- 02Calibrate
Capture representative demonstrations
- 03Produce
Label states and transitions
- 04Learn
Evaluate complete behavior