Ticket Resale Platform

Built solo — every line, every deployment, every 2am bug. Here's what each piece does and why it exists:

• ML Fraud Detection — every listing hits a Scikit-learn model running as its own Flask microservice, scoring price deviation, account age, and daily listing volume before it ever goes public. High-risk listings get flagged before buyers see them, not after.
• Real-Time via Socket.io — the second a ticket sells, a WebSocket event wipes that listing from every connected screen instantly. No polling, no risk of selling the same ticket twice.
• Two Databases, Split by Purpose — MongoDB Atlas handles live transactional reads/writes; PostgreSQL on Supabase absorbs heavy analytics queries separately, so reporting never slows down the live marketplace.
• JWT + QR Verification — every purchase generates a single-use QR cryptographically tied to that buyer and ticket. Scanned once, then dead — kills the screenshot-and-resell scam.
• 3 Services, Deployed Independently — React, Node API, and the Flask ML service each deploy on their own. Retrain the model without touching the live app; if ML goes down, the marketplace just skips the fraud check instead of crashing.

Three independently deployed, independently scalable services. Real-time updates that eliminate double-booking. An automated fraud layer that catches high-risk listings before a human ever reviews them — and because it runs as its own microservice, scoring happens asynchronously without blocking a listing from loading.

Beyond the app itself, this is where I learned to actually operate a multi-service production system solo — coordinating deploys across three platforms, handling a service going down without taking the whole app with it, debugging real-time bugs that only surface with multiple concurrent users.

It's live, it's public, and I'm actively open to feedback — and to SDE / full-stack roles.