Continuous on-body sensing with classification on device and multi-day battery life.
A wearable band that senses activity, heart rate and skin temperature, classifies context on device, and syncs batched results over BLE. Battery life is treated as the primary product specification rather than an outcome.
Every feature request translated directly into average current. The team needed a way to say yes or no to features with evidence, not opinion, and lab sensor data did not transfer to real wear across different body types.
We built a measured current budget in the first week and held every feature against it. Feature extraction moved on device so raw signal never streams, and a signal-quality gate rejects unusable windows before inference runs at all.
An ultra-low-power MCU with a PPG front end, IMU and skin temperature sensor on a rigid-flex board inside a sealed band.
A TinyML activity classifier and signal-quality gate running in fixed point on device, with longitudinal trend analysis in the cloud.
Zephyr firmware with a flash-backed store-and-forward buffer, plus iOS and Android apps with background sync and health platform integration.
Hardware
AI
Embedded
Apps