TrackSheet
Flutter, Firebase, BLoC, REST APIs, Hive, Shorebird, Golang. Drove engagement growth to 1.3K daily and 5K monthly active users. Implemented GitHub Actions, Fastlane, and Shorebird CI/CD, reducing release time by more than 50%. Maintained 0.281% crashes and 0.069% ANRs across 2.21K active devices. Designed PostgreSQL and MongoDB backends with Redis caching and JWT authentication, achieving 200 ms latency. Handled 365K requests in 5 days with a 99.5% success rate on Cloud Run and Docker with monitoring.
CodeCollab (Backend)
Golang, WebRTC (Pion), AWS Lambda, Prometheus, and Grafana. Built a real-time linting backend using WebSockets and Supabase JWT, with 60 rpm rate limiting and structured error handling. Orchestrated AWS Lambda linting for more than five languages through a unified WebSocket contract. Implemented Discord-like WebRTC voice calls and rooms with STUN/TURN, scaling rooms to more than 50 participants. Added Prometheus, Grafana, and Loki observability with health and metrics endpoints.
Code-Collab Backend
Ever tried pair programming remotely and thought "there has to be a better way"? That's what sparked CodeCollab. I built the backend infrastructure in Golang to power real-time collaborative coding. The challenge: let multiple developers code together, talk to each other, and get instant feedback - all in real-time. What I Built: Real-time Sync: WebSocket connections handle live code synchronization. When someone types, everyone sees it instantly. Added JWT auth with Supabase and rate limiting (60 req/min) to keep things secure and spam-free. Voice Chat: Integrated WebRTC using Pion for an SFU (Selective Forwarding Unit). Set up STUN/TURN servers so it works even behind corporate firewalls. Handles 50+ people in a room - though honestly, who needs that many people coding together? Smart Linting: Instead of expensive server-side processing, I use AWS Lambda functions that spin up on-demand. Supports Python, JavaScript, Go, Java, and more. Everything flows through a single /ws WebSocket endpoint keeping the API clean. Observability: Added Prometheus, Grafana, and Loki for monitoring. Now I can actually debug at 3 AM when things break. Health checks, metrics, logs - the full setup. Deployment: Containerized with Docker Compose, deployed on EC2. Implemented graceful shutdowns because losing work on server restarts sucks. Tech: Golang, WebSockets, WebRTC (Pion), AWS Lambda, PostgreSQL, Docker, Prometheus/Grafana. Great learning experience in real-time systems and peer-to-peer communication. Plus I finally understand what SFU actually means Demo: https://code-collab-iota-eight.vercel.app/ Code: https://github.com/SRAYANSH-GUPTA/code-collab-backend
