A sensor under a mattress sees a faint, noisy waveform — body movement, breathing and heartbeat all superimposed. Turning that into trustworthy heart rate, respiration and HRV is where the engineering lives.
Clean at the edge, decide in the cloud
Firmware on the device (C++) handles sampling and a first pass of filtering, then ships compact frames over MQTT. Heavy spectral work — FFT, peak detection, HRV computation — runs server-side where we can iterate on algorithms without reflashing hardware.
Separate the signals before you trust them
Heart rate (HR), RR interval, stroke volume and HRV each demand different processing windows. We isolate the cardiac band from respiration and motion before deriving any metric, because a number computed from contaminated signal is worse than no number at all.
Treat health data like the liability it is
Telemetry is sensitive by default: encrypted in transit and at rest, access-controlled, and never presented as a certified medical device unless it is one. Traceability lets us reconstruct how any reading was produced.
Built on C++, MQTT, Node.js, React and PostgreSQL/MongoDB over AWS — the same end-to-end discipline that makes IoT products survive contact with the real world.