Jeye

The problem

Drivers have to keep broadcasting their location while the app is backgrounded and the network is patchy, without flooding the server, and without a driver looking offline just because they switched apps.

Key decisions

1. 01

   Drivers stream location over Socket.IO. Every ping is validated, then throttled by both time and distance moved, so a phone in a moving car doesn’t flood the backend.

2. 02

   The last known position lives in Redis with a 60-second expiry and is pushed live to that branch’s operations dashboard and to the customer tracking the order.

3. 03

   Being reachable and being online are tracked separately. A 90-second expiry, refreshed by heartbeats and location pings, means a driver who backgrounds the app stays online, while a crashed app drops off by itself, even across a Redis restart.

4. 04

   The delivery fee comes from real driving distance plus uphill elevation, priced per region by matching the customer’s location against map boundaries, in USD or LBP, with routes cached so the maps APIs aren’t called twice.

5. 05

   Background scans flag stuck orders, orders with no driver and uncollected cash, and accepting a delivery runs in a strict database transaction that re-checks the driver’s cash limit, so two taps can never double-accept an order.

Outcome

From first commit to launch in about four months, as tech lead: the backend and both Flutter apps, published on the App Store and Google Play under my name. Arabic and English.