Logistics Automation System
Designed and delivered an end-to-end logistics automation platform replacing manual dispatch operations with intelligent route optimization.
A freight and supply-chain operator dispatched hundreds of routes a day by phone and spreadsheet. We replaced that with a supervised automation platform: the system proposes optimized daily plans, dispatchers review and adjust, and drivers receive live manifests — turning tribal knowledge into an asset the business owns.
A freight operator dispatched hundreds of daily routes by phone and spreadsheet. Dispatchers were the bottleneck: route decisions depended on individual experience, and a single sick day degraded the whole operation.
An automation platform that turned dispatch into a supervised process: constraint-based route optimization proposes daily plans, dispatchers review and adjust, and drivers receive live manifests. Exceptions — breakdowns, cancellations — reroute in minutes instead of hours.
Event-driven services on Node.js with Redis streams carrying fleet telemetry, and PostgreSQL as the system of record. The optimizer runs as an isolated worker pool so heavy computation never blocks live operations.
Key Features
Constraint-based route planning
Daily plans generated against real constraints — vehicle capacity, time windows, driver hours — proposed for human review, not blind execution.
Live manifests for drivers
Each driver gets a current route on their device; changes propagate in seconds, not the next morning's briefing.
Exception rerouting
Breakdowns and cancellations trigger a fresh proposal in minutes, keeping the day on track instead of unravelling it.
Dispatcher control room
A single operational view of the whole fleet, so a planner covers what used to take three.
Technical Decisions
Supervised automation, not autopilot
Full autonomy was technically possible but not trusted. We put the optimizer's proposal in front of a human who approves or adjusts — the design the team actually adopted.
Isolated optimizer worker pool
Route optimization is compute-heavy. It runs in an isolated worker pool so a big recalculation never blocks live dispatch operations.
Event-driven core on Redis streams
Fleet telemetry flows through Redis streams into event-driven services, with PostgreSQL as the system of record — fast where it must be, durable where it counts.
35%
operating cost reduction
99.97%
uptime across two peak seasons
3×
dispatch throughput per planner
The operator absorbed two seasonal peaks without hiring additional dispatchers, and route knowledge moved from individual heads into a system the business owns.
Development Process
Discovery
Shadowing dispatchers to encode the decisions they made from experience into explicit, reviewable constraints.
Optimizer prototype
An early planning engine validated against historical routes before any driver saw it.
Supervised rollout
Dispatchers ran the system alongside their old process until they trusted the proposals enough to lead with them.
Peak-season hardening
Load-tested and tuned against two full seasonal peaks before the manual fallback was retired.
Future Roadmap
- Predictive demand planning
- Carbon-per-route reporting
- Customer delivery-window portal
- Driver mobile app v2