I built a live Legal Assistant Chatbot with Python, Claude and OpenAI APIs, and a FAISS-backed RAG pipeline that grounds responses in legal source documents. I iterated on prompts and evaluations to make answers consistent and traceable, then deployed the chatbot on Streamlit Community Cloud.
As a Research Intern (Machine Learning) at IIT Bhubaneswar, I designed experiments for adaptive parameter selection in federated learning using Deep Reinforcement Learning. I also developed a modular object-detection and retrieval pipeline, and evaluated sentiment classifiers using macro F1 and F-beta scoring.

