Counsel.ai — AI-powered Legal RAG system (project)
Built an AI-powered Legal RAG system to analyze complex legal documents and generate context-aware legal advice. Implemented retrieval and dynamic context assembly using a vector search pipeline to ground responses in provided sources. Developed an application workflow that supports real-time question answering from uploaded PDFs while reducing hallucinations through prompt engineering. • Used DeepSeek-R1 embeddings with a Groq-based inference setup. • Built a LangChain + FAISS semantic search and retrieval pipeline. • Created a Streamlit interface for PDF upload and grounded legal Q&A. • Engineered custom prompts to improve accuracy and source grounding.