Research-grade wearable data
Serious science from the watch on your wrist.
The challenge
Consumer wearables are full of useful sensors, but operating systems throttle background apps, clocks drift and batteries drain. The data is rarely clean enough for research.
Our approach
A collection system that records continuously in the background, checks and repairs data quality, and validates sensor accuracy with an industrial robot arm that replays real human motion.
Highlights
- ~1.4% battery used over 14 hours of continuous recording
- 99.99%+ fewer processor wake-ups thanks to smart batching
- 94%+ accuracy in on-device walking-speed classification, fully offline
- Side-by-side accuracy testing across device brands and models, under identical conditions
Built on
Robotic sensor validation