AI Security Log Analyst — RAG pipeline for threat questions.
Built end-to-end for the MSc placement at TU Shannon: raw security logs are ingested by a Python Watchdog file watcher, chunked and embedded with SentenceTransformers (all-MiniLM-L6-v2) into ChromaDB, then answered in natural language by chaining semantic retrieval with Ollama / LLaMA 3.2 through LangChain LCEL — across millions of log events.
The 5-component stack (ingestion, ChromaDB, FastAPI, Ollama, React + Vite) ships as a fully offline, privacy-first Docker Compose deployment with a Groq API cloud fallback, validated with Pytest unit tests and Locust load testing across 3 API endpoints.