Spice Garden — Restaurant Management Platform

A production system built for a real restaurant, from first request to a running kitchen.

The problem

A friend running a restaurant needed to move off paper order slips and manual billing. The ask: real-time order tracking from table to kitchen, GST-compliant billing, and a way to see how the business was actually performing — all for a live restaurant, not a demo.

What I built

• A Kitchen Display System and waiter-facing orders panel, with orders moving in real time via Socket.io as soon as they're placed.
• GST-aware billing built to match how the restaurant actually invoices customers.
• Payment analytics and operational dashboards so the owner can see sales and order trends without digging through receipts.
• A PostgreSQL (Supabase) data layer with a Node.js/Express API and a React frontend, deployed on Railway (backend) and Vercel (frontend).

Problems I had to actually debug in production

• Railway–Supabase connectivity: intermittent connection failures traced to an IPv6/IPv4 mismatch between the two platforms — fixed by forcing the correct connection mode.
• CORS and QR-code ordering flow: customers scan a table QR code to order directly, which meant getting cross-origin requests and dynamic table sessions working reliably from phones on restaurant Wi-Fi.
• Timezone bugs across the stack: order timestamps were drifting because SQL, the frontend's date handling, and node-postgres were each interpreting time differently — normalized all three to IST consistently.

Why it matters

This wasn't a portfolio dataset or a tutorial clone — it's software a real restaurant runs on, which meant the bugs were real too: production connectivity failures, real customers hitting the QR ordering flow, and timestamps that had to be correct for actual billing. Debugging those end-to-end, across infrastructure, network, and data layers, is the part that doesn't show up in a screenshot.