24/7 operations don't accept downtime. Not the team, not the system.
Traustar coordinates transport fleets and logistics chain. The system runs every day, every hour, no pause. Any minute of downtime has a measurable cost in money and in client trust.
Traustar needed the backbone of a 24/7 operation.
Traustar operates a fleet of [N] vehicles coordinating routes, load assignments, fuel telemetry, and real-time customer notifications. Every route change, every road incident, every driver rest hour — everything syncs with the central system. If the system goes down, vehicles don't stop, but clients lose visibility and coordinators work blind.
When we arrived, the previous system had visibility but no scale: every time a new fleet was added, part of the backend had to be duplicated manually. Operations had grown faster than the architecture supporting them.
What we understood in discovery.
We spent the first three weeks in operations — not in an office, in operations. On vehicles when possible, watching the coordinator's dashboard live, listening to client calls asking for ETAs.
What we discovered: the problem wasn't system performance. It was that the data model assumed the fleet was fixed, and reality was that the fleet was reconfigured weekly. Any optimization on top of the old model would be temporary.
Final architecture.
NestJS on PostgreSQL with regional sharding. Operations are divided regionally — the system is too.
Vehicle data ingestion via MQTT, processing in Cloudflare Workers, deferred persistence in Postgres + historical table in R2.
Integration with optimization engine (OR-tools) for real-time route recalculation when incidents occur.
Next.js with real-time dashboard for coordinators. Maps with Mapbox. Push updates via WebSockets.
React Native with offline-first sync. Vehicles lose signal in rural areas — the system keeps working locally.
NestJS, PostgreSQL, Redis, MQTT, Cloudflare Workers, R2, React Native, Mapbox.
Production metrics.
PROVISIONALReplace with verifiable real metrics before publishing.
“Migrating to FastNet cut our deploy time from hours to minutes. But more importantly: when something breaks, it's no longer a lottery figuring out who understands what.”