Aman Vijay

Built a real-time notification service from scratch using FastAPI, Redis steams, and Supabase — chose Redis stream with sepearate channels and DLQ, and shipped it to handle 500 concurrent sessions at sub-200ms latency. Worked on an internal RAG pipeline using Python and FastAPI: built ingestion and embedding jobs, improved retrieval relevance, and added logging and basic metrics to make debugging retrieval failures faster.

Set up Supabase RLS for authentication and role-based access control across admin, co-admin, and security roles — applied JWT middleware for route-level enforcement during concurrent usage. Migrated frontend state management to Zustand, replacing scattered Context usage restructured stores into independent slices, which made state updates consistent and reduced debugging time across the application.

Led backend design for core services using Node.js/Express — built normalized MongoDB schemas, input validation, and authentication with Google OAuth 2.0 + Passport.js; added route-level access control through custom middleware. Set up the CI/CD pipeline from scratch on AWS EC2 with Docker and NGINX — automated deployments and rollbacks via GitHub Actions, cutting manual release effort and maintaining 99.9% uptime post-launch.

Integrated OpenAI GPT APIs into the MERN stack using WebSocket-based streaming instead of REST polling, delivering real-time conversational responses with lower perceived latency for end users. Tracked down and fixed frontend performance bottlenecks using Lighthouse — applied lazy-loading and React.memo to cut page load times by 40% and improve user engagement metrics by 25%.