Project Details
Home/Project Details
VIC Alert
Project description
An assistive alert system whose first requirement is that it works when everything else does not
The Challenge
An assistive alert system is only useful in the situation it was designed for, which is by definition the situation where things are going wrong. A server that is unavailable during a network problem, or that drops a device's alert because it arrived twice, has failed the person it exists for.
Devices are low-power and connections are unreliable, so the ingest path has to tolerate duplicate reports, out-of-order arrival and long gaps — and the server cannot assume a device is online to ask it anything.
The Solution
Design the ingest as idempotent and the alert pipeline as append-only: a device reports state, the server records what it received and when, and alerts are raised from that record rather than from the live connection. Nothing is lost because a socket died.
Skills and deliverables
Features delivered
The hard part
Deciding what to do with silence. A device that has stopped reporting may be switched off, out of range, or the emergency itself — so the design distinguishes "no data" from "fine" and escalates on the former rather than treating it as normal.
The Impact
The server component built: device ingest, environmental readings, SOS alerts and the escalation rules around missing data. It is a backend service with no front end hosted, so the cover describes it as exactly that.
