Designed and deployed a multi-agent AI system for automated patent analysis, executing novelty detection, infringement mapping, and Freedom-to-Operate (FTO) workflows. Built a high-performance RAG pipeline over 7,000+ patent embeddings using FAISS and sentence-transformers, enabling source-backed reasoning over large-scale patent data. Architected a 6-agent asynchronous pipeline (Search, Filter, Novelty, Claim Mapping, Legal Reasoning, Citation) using Python asyncio, reducing latency from ~30s to <3s (12x speedup). Integrated Groq-accelerated LLaMA 3.1 inference with strict JSON schema validation to eliminate hallucinations and ensure deterministic outputs. Designed system scalable to millions of patent documents with autonomous retrieval + reasoning execution. Tech Stack: Python, FastAPI, FAISS, LLaMA 3.1, Groq, Streamlit, asyncio, Docker, Sentence-Transformers