We've seen plenty of cool generative AI coding demos lately, but enterprise developers looking to adopt generative AI face challenges such as scalability and contextual awareness. Retrieval Augmented Generation (RAG) can help bridge this gap by indexing knowledge bases and retrieving relevant code snippets. However, implementing RAG with large codebases requires intelligent chunking strategies that respect the structure of the code, maintain context in chunks, and handle different file types. Enhancing embeddings with natural language descriptions further improves retrieval for queries. Advanced retrieval techniques, such as two-stage search and repo-level filtering, reduce noise and improve relevance. By developing a scalable architecture and evaluating performance using multi-faceted metrics, we can effectively navigate and leverage the vast knowledge contained in enterprise-scale codebases.