Ajay Garg@ajaygarg200r81231

About me

I started where most CS students do — building full-stack CRUD apps to learn how the pieces fit together. But I got more interested in what happens behind the UI: how authentication actually works at the token level, how payment flows handle failure states, how WebSocket connections scale under load. That led me deeper into backend architecture — building real auth systems with hash-stored single-use refresh tokens, integrating Agora SDK for real-time video, and deploying services across Railway and Vercel with proper CI/CD pipelines. More recently, I've been building agentic AI systems — not wrappers around ChatGPT, but actual multi-service architectures where an Orchestrator routes tasks to specialized Planner, Memory, and Executor services, each running in its own Docker container. The system (VEDA) uses LLM-driven task decomposition, vector-based memory retrieval via FAISS, and automated deployment through GitHub Actions. I also co-authored research on self-healing cloud infrastructure, combining autoencoder-based anomaly detection with automated remediation — the kind of system that detects a memory leak before the on-call engineer's pager fires.

Socials