SUJEET
KUMAR

Built a multi-agent RAG system using LangGraph that routes queries across 6 specialist AI agents: Q&A, bug root-cause analysis, diagram generation, auto documentation, impact analysis, and repo onboarding. Designed an ingestion pipeline that clones GitHub repos, parses 15+ file types, chunks and embeds code using HuggingFace sentence-transformers, and stores it in ChromaDB for semantic retrieval. Implemented auto Mermaid diagram generation from source code using Google Gemini 2.5 Flash and deployed it on Streamlit Cloud.

Developed a full-stack monorepo application integrating a Daily Notes Tracker and a secured Password Management System. Integrated an AI engine for one-click summarization, implemented secure OAuth authentication and bcrypt hashing, added database-level authorization for note sharing with automated 3-day TTL expiration, handled 100k+ concurrent users during testing, and containerized everything with Docker.

Developed an AI-powered chatbot with secure JWT-based authentication and authorization. Implemented token issuance during login and middleware-based validation to protect backend routes before invoking LLM APIs. Enforced access control by rejecting invalid or expired tokens, preventing unauthorized API usage and ensuring secure interaction with AI services.