𝗜𝗻𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻 𝗥𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 | 𝗖𝗦𝟲𝟬𝟬𝟵𝟮 | 𝗕𝗼𝗼𝗹-𝗦𝗲𝗮𝗿𝗰𝗵, 𝗥𝗮𝗻𝗸𝗲𝗿, 𝗪𝗼𝗿𝗱𝗡𝗲𝘁 𝗦𝘂𝗺𝗺𝗮𝗿𝗶𝘇𝗲𝗿

A project demonstrating the ability to utilize Agentic AI workflows to perform analysis and discovery on aspects of US Healthcare Taxation Laws. Utilizes an underlying RAG framework for Response Generation. Also includes a comparison of LlamaIndex vs a custom framework that makes use of retrieval strategies like Hyde and Multi-Query Transformations

Compare 4 RAG query transformation techniques side by side — Rewrite, Multi-Query, Step-Back and HyDE

Advanced RAG system for tech docs. Powered by Qdrant (dense vector), Neo4j (GraphRAG), BM25 (lexical), Cross-Encoder reranking, and query transformations.