Back to the blog
Architecture

How Your Watch Talks to the Cloud

Follow a single alert from the sensor on your wrist to a caregiver's phone — the full path it travels, step by step.

Every alert LifeLyne sends starts with something simple: your watch noticing that something has changed. Getting from that first flicker of data to a caregiver's phone — reliably, in seconds, without crying wolf — takes several handoffs between the device, the app, and the cloud.

Scroll to follow the alert
1

On the Wrist

LifeLyne currently runs as an app on off-the-shelf Android smartwatches, like the Samsung Galaxy Watch — no custom hardware to design or wait on. The watch continuously samples heart rate, accelerometer, gyroscope, electrodermal activity, and skin temperature in the background.

2

On-Device Intelligence

A lightweight machine learning model runs directly on the watch's chip, scanning sensor data in real time for patterns consistent with a fall or seizure — entirely offline, with no round trip to the cloud needed to notice something's wrong.

3

Encrypted Handoff

The moment a pattern is flagged, the watch alerts the paired phone over an encrypted Bluetooth Low Energy connection. That connection is health-checked continuously in the background, even when the screen is off, with automatic reconnection if it ever drops.

4

A Careful Second Look

The app relays the event to LifeLyne's cloud backend, where a deeper model reviews the full picture before deciding whether to escalate. This second pass — a heavier model with more context than the watch alone can hold — is what keeps false alarms down.

5

The SOS Window

Before anyone is contacted, the app shows a short on-screen countdown. If it's a false alarm — a hard workout, a dropped phone — the wearer can cancel it right there. If it goes unanswered, the app proceeds automatically.

6

Reaching Help

Caregivers are notified by SMS and phone call, with the wearer's location included automatically. If the phone has no signal, it falls back to the phone's own dialer directly — the alert still goes out.

WearableSensors + on-device MLMobile AppBLE link + SOS countdownCloud BackendML confirmationCaregiverSMS + call + location

Behind the Curtain

The cloud side is built to stay out of the way — scaling automatically, and treating health data as something to isolate, not just store.

Serverless & auto-scaling on AWS
Isolated, encrypted database — no public access
Infrastructure managed as code

That's the whole trip —
watch to caregiver, in seconds.

Curious what it looks like in the app itself?

See It Work
Continue ReadingThe Sensor Science Behind DetectionBack to the blog