Module assessment

1.

A clinical Retrieval-Augmented Generation (RAG) agent must retrieve drug information by exact National Drug Code (NDC) code while also matching synonyms like 'myocardial infarction' for 'heart attack.' Which Azure AI Search configuration is most appropriate?

Pure vector search using text-embedding-3-large to capture all query semantics.

Hybrid search combining BM25 keyword search and vector search with Reciprocal Rank Fusion (RRF) fusion, weighted toward BM25 for formulary queries and toward vector for guideline queries.

Keyword-only BM25 search to ensure exact clinical term matching throughout the pipeline.

2.

After initial retrieval returns 50 candidate documents, your clinical RAG pipeline needs the highest possible precision for the final three documents injected into the agent context. Which reranking approach delivers this?

Azure AI Search built-in semantic ranking applied to all 50 results using the configured semantic configuration.

Cross-encoder reranking to reduce 50 to 10 results, followed by an LLM-as-reranker prompt to select the final 3 from those 10.

Rerunning the vector search with a higher similarity threshold to filter the 50 results down to three candidates automatically.

3.

Your Retrieval-Augmented Generation (RAG) pipeline uses fixed-size chunks of 512 tokens. Clinical guidelines sometimes split mid-criteria, causing the agent to miss key diagnostic conditions that span chunk boundaries. What chunking strategy resolves this?

Increase fixed chunk size to 2,048 tokens to ensure complete sections are always included in a single chunk.

Hierarchical (parent-child) chunking: index small chunks for precise retrieval and return the parent section when a child chunk is matched.

Semantic chunking using embedding similarity to detect topic boundaries, then splitting at those boundaries.

Feedback