Transport Tracker

Transport data arrives from several operators, each with a different feed format, update frequency and definition of "on time". Naively merging them produces a view that flickers, contradicts itself, and loses trust the moment a predicted arrival jumps backwards.

Transport Tracker consumes multiple real-time feeds, normalises them to a common schema, and maintains a single reconciled state of where every vehicle is and when it will arrive. A streaming layer pushes only the deltas to clients, so the map stays smooth on a phone on a slow connection.

The hard part is not ingesting feeds, it is deciding which source to believe when they disagree. The system keeps a confidence-weighted estimate per vehicle and damps predictions so arrivals never jump erratically — a slightly stale but stable number beats a precise one that jitters.

Real-time engineering end to end: multi-source ingestion, state reconciliation, streaming deltas and a UI tuned for trust under noisy data.