About the Client
HattaFoodHub is a food delivery startup connecting restaurants, couriers, and customers across multiple cities. They entered the market with a strong brand and loyal restaurant partners, but their technology stack could not keep up with growth.
The Problem
HattaFoodHub had a working MVP — but it was built on a monolithic Laravel backend tied to a single MySQL database. It worked fine for 50 orders an hour. At 200 orders, it slowed down. At 500, it crashed. Lunch and dinner peaks were a daily gamble: would the platform stay up today?
- Orders took 8–12 seconds to confirm during peak hours
- Driver dispatching had no real-time component — couriers refreshed their screens manually
- Restaurant dashboards showed stale inventory — menus would sell items that were out of stock
- No automated retry logic for failed payments — customers were charged without orders being created
What We Built
We rebuilt the entire platform with a microservices architecture:
- Order Management Service — handles the lifecycle from cart creation to delivery confirmation, with idempotency keys so double-charges are impossible
- Dispatch Engine — a real-time WebSocket service that matches couriers to orders based on proximity, load, and vehicle type, with automatic reassignment if a courier rejects or times out
- Payment Gateway — unified integration with Stripe, with retry logic, webhook reconciliation, and automatic refunds for cancelled orders
- Restaurant Dashboard — a React frontend that shows live inventory sync, incoming orders with audio alerts, and analytics on peak hours and popular items
- Customer App — React Native app with live GPS tracking of couriers and estimated delivery times recalculated every 30 seconds
Technical Approach
We containerized each service with Docker and orchestrated them on AWS ECS with auto-scaling. Redis handled session caching and rate limiting. PostgreSQL became the transactional database. The entire stack deployed through a GitHub Actions CI/CD pipeline with blue-green deployments — meaning updates went live with zero downtime.
“I can finally sleep during dinner rush. The platform just handles it.”
The Results
The new platform launched without a single outage. Peak-hour order confirmation dropped from 8–12 seconds to under 800 milliseconds. Repeat customers increased by 40% in the first two months — users came back because the app was fast and reliable.