I built an Agentic AI Banking Fraud Investigation System that coordinates specialized agents to investigate suspicious financial transactions. It retrieves transaction data from PostgreSQL and combines findings into structured reports using Groq Llama 3.1.
For Think Python AI Assistant, I built a hybrid RAG application that answers questions from the Think Python, Second Edition knowledge base. It uses semantic retrieval and vector search, with Tavily web search as a fallback and source-aware responses with PDF citations.
I also built Automora AI, a conversational lead-generation system that uses natural question-and-answer interactions to qualify leads and extract structured information. My Deepfake Detection System compared transfer-learning models for real-versus-fake face classification and tested robustness to image changes.
