Velora

Delivery fleets juggle dozens of stops, vehicle capacities, and time windows. Planning routes by hand is slow and rarely optimal, and naïve solvers block while they compute — no good for an interactive dashboard.

I built a scalable VRP engine on FastAPI + Celery + Redis that runs optimization jobs concurrently with VROOM, so the UI never blocks. Constraint-aware scheduling handles vehicle capacity, time windows, and multi-stop routing; OSRM + PostgreSQL power route analytics and optimization logs.

A Next.js dashboard (Zustand for fleet state, Google Maps for visualization) submits optimization jobs to a FastAPI API, which enqueues them on Celery workers backed by Redis. VROOM solves the VRP while OSRM provides the road network; results and metrics persist to PostgreSQL. The client polls job status with TanStack React Query for real-time updates.

1. 01Keeping the API responsive while long-running solves happen — solved with async job queues and status polling.
2. 02Modeling real-world constraints (capacity, time windows, multi-stop) without exploding solve time.
3. 03Streaming live optimization status to the dashboard without hammering the server.

• ✓Bronze Medal at Kriti, IIT Guwahati's inter-hostel technical competition.
• ✓Concurrent optimization workflows that stay responsive under load.
• ✓Clear route analytics and performance metrics for every run.