Custom data
Data collected to your spec, at the scale you need.
When the catalog isn't enough, we design, capture and annotate new datasets end to end — with the same consent, provenance and QA standards.
01PHYS
Physical AI programs
Egocentric and exocentric video, teleoperation and motion capture in the environments your robots will work in.
- Capture
- Head-mounted, wrist, multi-view, RGB-D, IMU, mocap
- Environments
- Homes, kitchens, warehouses, retail, factories, outdoors
- Annotation
- Hand pose, actions, objects, contact, language
- Typical scale
- 1,000–50,000 hours
02LLM
LLM programs
Expert-authored SFT, preference and agentic data in the domains, languages and formats your roadmap requires.
- Data types
- SFT, preference pairs, rubrics, agent trajectories, evals
- Experts
- PhD and professional networks across 24+ domains
- Languages
- English plus 20+ languages on request
- Typical scale
- 10k–1M examples
03Process
A pilot in weeks, production in months.
Every program starts with a paid pilot: we agree on the spec, deliver a pilot batch with quality metrics, and scale once you sign off.
- 01
Capture
A vetted global network of contributors, operators and domain experts records data on calibrated devices — in homes, workplaces and labs.
Head-mounted cameras · teleop rigs · expert authors
- 02
Annotate
Trained annotators and model-assisted tooling add hand pose, action segments, language, preferences and rationales.
Hand pose · actions · narration · rubrics
- 03
Verify
Multi-pass human review, automated checks and contamination screens. Every batch ships with measured quality metrics.
Two-pass QA · agreement metrics · audits
- 04
Deliver
Clean rights and provenance, delivered in the formats your training stack already reads — to your bucket, on your schedule.
LeRobot · RLDS · WebDataset · Parquet · JSONL
Custom programs
Let’s scope your dataset.
Share your target tasks, modalities and volume. We reply within one business day with matching datasets, samples and a quote.