Emergency Button Network

Emergency Button Network

This emergency alert system now covers roughly 100 public facility locations. It started as a single-site proof of concept. I designed and built it end to end, then carried it through to full deployment.

Challenge

A panic button is a system you hope never fires. That flips the usual engineering problem on its head. Uptime can't be proven by traffic: a healthy button and a dead one look identical when nobody presses anything. Site operators needed to know a button was still alive before an emergency, not after.

And when a press did happen, it needed to reach the right responder within seconds. It couldn't sit unread in a channel nobody was watching that day.

Solution

The backend runs on a supervised heartbeat. Every field gateway reports in on a fixed interval. A background worker sweeps the fleet continuously. A silent site moves to warning after one threshold, then to offline after a second.

Presence and alarms travel as separate signals. A heartbeat that just repeats "still here" never creates an incident, so operators aren't trained to tune out real ones. Alerts route by role and by area. An escalation chain moves an unacknowledged press to a supervisor, then to a central desk on a timer.

Every press, state change, and notification lands in an append-only log. Buttons carry an armed or disarmed state. An accidental press gets recorded without paging anyone.

Impact

The network now runs across roughly a hundred distributed sites, up from one, with liveness checked continuously instead of assumed. A dead button surfaces as a warning within about a minute, rather than staying silent until someone needs it. False alarms get logged and closed out without triggering escalation.

That keeps the alert channel worth watching instead of one more thing to mute. It's the same discipline behind scaling a panic button fleet to 100 locations.