Optimizes RAG pipeline chunk sizes, overlap, and embedding model selection for maximum retrieval quality.