Built "MedQA," a RAG-powered medical Q&A system for accurate & relevant answers, utilizing 100 Medical PDFs.
● Accomplished a 35% improvement in medically sound and natural language responses compared to traditional methods by utilizing the "Phi-2" pre-trained language model.
● Employed FAISS, a high-performance vector search library, for efficient text retrieval in MedQA's medical knowledge base, giving lightning-fast text search within the PDFs. Combined with iterative refinement, responses gained 40% relevance, pinpointing user needs.
● Streamlined development by 30% with Langchain's modular framework, enabling 2x faster prototyping of different retrieval and LLM combinations.
