Ground truth from the real world, in sync.
We capture continuous, time-aligned data on how people move, rest and live in real spaces. It's built for teams training physical AI and running the next generation of health research.
Models that act in the physical world need data from the physical world. Almost none of it is captured continuously, aligned in time, and labeled well enough to train on.
Lab studies are short and staged. Consumer devices each keep their own clock and format. Teams end up spending months stitching streams together before they can ask a single question. We do that work before the data reaches you.
One record, not a pile of streams.
Passive, in real spaces
Unscripted data from real homes and rooms over days and weeks, not minutes in a lab.
Every channel, one clock
Body, motion and environment land on a shared timeline, so cause and effect line up.
Ready to train on
A consistent schema, event labels and documentation that your pipeline can load on day one.
Camera-free by design. Our capture never records video or images of the people or spaces we measure.
Teams that need the real thing.
Robotics & world models
Long-horizon human activity in real environments, aligned across modalities, for pre-training and evaluation.
Wearables & digital biomarkers
Reference-grade context to validate new sensors and algorithms against what actually happened.
Trials & longitudinal studies
Continuous at-home measurement between visits, collected under clear consent and data protocols.
Start with a pilot.
Scope, protocols and pricing are shared under NDA.
A dataset built around your question
A fixed-scope capture designed with your research team, delivered as a documented, analysis-ready dataset.
- Length
- 4–8 weeks
- You get
- Dataset, schema, protocol report
- Best for
- Validation and feasibility
A standing stream you can build on
Recurring access to new sessions in the same format, licensed for training and evaluation.
- Cadence
- Monthly or quarterly drops
- You get
- Versioned releases, usage license
- Best for
- Model training and benchmarks
Built for review boards and legal teams.
Consent first
Every participant opts in knowingly and can withdraw at any time.
De-identified by default
Datasets ship without direct identifiers, with documented handling steps.
IRB-ready protocols
Study designs written to go through institutional review without rework.
Clear licensing
Plain terms on what you can train, publish and keep.
Request a pilot brief.
Tell us what you're building. We'll reply within two business days.