Personal RAG Assistant

I built a fully working chatbot that lets recruiters and employers ask natural questions about my background, experience, projects, and technical decisions. Instead of guessing or hallucinating, the bot searches through my actual resume, project notes, and write‑ups before answering. This is called Retrieval‑Augmented Generation, or RAG, and I built the entire system from scratch.

The project has two versions. The local version runs on my own computer and shows a complete AI engineering workflow. It ingests documents in Markdown, text, PDF, and Word formats, breaks them into chunks, creates embeddings using Ollama, stores them in ChromaDB, and then retrieves the most relevant chunks to answer questions with a local language model. The backend is built with Python and FastAPI, and the frontend is a chat interface built with React and TypeScript. This version proves I understand the full RAG pipeline, not just how to call an API.

The hosted version is a lightweight public demo deployed on Cloudflare Workers. It exports the same knowledge base into a JSON file and uses Workers AI to generate answers. This shows I can take an AI prototype and make it accessible online with real deployment constraints, including managing free tier limits and environment configuration.

When a recruiter visits the hosted chatbot, they can ask things like "What cloud experience does he have?" or "Has he worked with vector databases?" and get accurate, sourced answers pulled directly from my resume, experience files, and project case studies. The bot never makes up information because it is grounded in my actual documents.

This project demonstrates that I can work with local LLMs and vector databases, build production‑ready APIs with FastAPI, create functional frontends with React and TypeScript, deploy to the cloud using Cloudflare Workers, and document complex systems clearly.