Fleet Telematics API Platform on NestJS

Telematics traffic is bursty: devices reconnect together after a network outage, and morning start-up brings a wave of ignition events. Devices publish over MQTT, and events flow into Kafka, which acts as a shock absorber between the devices and everything else. Consumers process at their own pace, so a burst from the field never slows the public API.

Time-series data lands in PostgreSQL with TimescaleDB, which keeps queries over long position histories fast, while Redis serves the latest known state of each vehicle for the most frequent requests.

• MQTT and Kafka pipeline with back-pressure
• TimescaleDB for long position histories
• Redis for the latest vehicle state

The public API is built with NestJS and TypeScript, with validation, pagination and error formats applied consistently across every endpoint. OpenAPI documentation is generated from the code itself, so the docs partners read always match the behaviour they get.

Versioning means partners can upgrade on their own timetable. Generated documentation cut partner integration time from weeks to days, because developers can explore endpoints, see example requests and responses and start building immediately.

Many partners prefer to be told when something happens rather than polling for it. The platform sends events such as completed trips and geofence exits to partner endpoints as signed webhooks, with automatic retries when an endpoint is unavailable.

Each partner can inspect a delivery log showing what was sent, when, and with what response, which turns "we never received it" from a support ticket into something partners can check for themselves.

• Signed payloads partners can verify
• Automatic retries with back-off
• Self-service delivery log for partners

Every API call is attributed to a key, which feeds billing and enforces rate limits fairly across customers. The provider can see which partners use which endpoints and plan capacity accordingly.

Just as important was the internal consistency. New endpoints follow one NestJS structure — modules, validation, logging, metrics — and services run in Docker on Kubernetes with rolling deployments. Any of the three teams can now work on any service, and releases ship several times a week without downtime.