EV Fleet Operations Platform & Telematics Ingestion

One unified telemetry stream from many vendors, powering a platform that runs an electric delivery fleet end to end.

Role: Architect, Lead Developer
When: 2025 – present
Where: Xtreme Thoughts SoftTech
Client: Electric last-mile delivery fleet

backend tests passing
client UAT cases passed
deploys needed to add a provider

The problem

Onboard riders, lease them vehicles, place them with delivery clients, track daily activity, monitor vehicle & battery health, then reconcile client invoices against rider and vendor payouts — with telemetry arriving in a different shape from every provider.

The system, drawn

Fig. 03One schema from many: webhooks → SQS → normaliser → history + live snapshotSimplified redrawing · illustrative data · real mechanism

How it works · 5 parts

01
Telematics ingestion
Provider webhooks authenticate with hashed bearer tokens; raw payloads are persisted, batched into AWS SQS, and a worker normalises each vendor's field and CAN mappings into one schema (dot paths, UTC timestamps, unmapped keys preserved). Output: de-duplicated history plus a live snapshot per vehicle per provider.
02
Operational resilience
DLQ classification auto-retries transient failures and surfaces the rest in a bulk retry panel. Per-provider health, uptime and failure counts drive alerts. New providers are onboarded from the admin UI with no deployment; a feature flag kept the cutover from vendor pull APIs reversible.
03
Realtime tracking
Rider positions stream over Socket.IO, throttled server-side to one write per 5s with a Redis route buffer, feeding live maps and route replay.
04
Domain breadth
Multi-step KYC, face + QR attendance with nightly auto-correction, vehicle master with insurance & EMI, full battery lifecycle, client invoicing, payouts and payroll.
05
Surfaces & delivery
Next.js ops console (Google Maps, Mapbox, deck.gl), React Native + Expo rider app, TypeScript/TypeORM ERP. Multi-stage Docker → ECR via GitHub Actions with per-environment and per-franchise pipelines.

Stack

Java Spring Boot
Node.js
Express
MongoDB
AWS SQS
Redis
Socket.IO
BullMQ
Next.js
React Native
Expo
TypeScript
TypeORM
PostgreSQL
S3
Docker
ECR
GitHub Actions

Next projectEnterprise Event & Budget Lifecycle PortalMedical-education events tracked from money down to evidence of delivery.

How it works · 5 parts

Telematics ingestion

Provider webhooks authenticate with hashed bearer tokens; raw payloads are persisted, batched into AWS SQS, and a worker normalises each vendor's field and CAN mappings into one schema (dot paths, UTC timestamps, unmapped keys preserved). Output: de-duplicated history plus a live snapshot per vehicle per provider.

Operational resilience

DLQ classification auto-retries transient failures and surfaces the rest in a bulk retry panel. Per-provider health, uptime and failure counts drive alerts. New providers are onboarded from the admin UI with no deployment; a feature flag kept the cutover from vendor pull APIs reversible.

Realtime tracking

Rider positions stream over Socket.IO, throttled server-side to one write per 5s with a Redis route buffer, feeding live maps and route replay.

Domain breadth

Multi-step KYC, face + QR attendance with nightly auto-correction, vehicle master with insurance & EMI, full battery lifecycle, client invoicing, payouts and payroll.

Surfaces & delivery

Next.js ops console (Google Maps, Mapbox, deck.gl), React Native + Expo rider app, TypeScript/TypeORM ERP. Multi-stage Docker → ECR via GitHub Actions with per-environment and per-franchise pipelines.