Interview Questions: The RAG Query Lifecycle
Practice answering like a builder who can explain, debug, and ship a RAG execution pipeline.
1. What is the RAG query lifecycle?
Strong answer: The journey from user question to retrieved evidence, grounded prompt, answer, citations, and evaluation.
Red flag: "It is just retrieval."
Follow-up: Which stage would you inspect first when an answer is wrong?
What this proves: You see RAG as a pipeline.
2. Why evaluate retrieval separately?
Strong answer: The model cannot use evidence that retrieval fails to provide.
Red flag: "Only the final answer matters."
Follow-up: What retrieval metric would you start with?
What this proves: You can debug stages independently.
3. Why do metadata filters matter?
Strong answer: They prevent semantically similar but wrong documents from being retrieved.
Red flag: "Embeddings handle everything."
Follow-up: Give one metadata filter for a refund bot.
What this proves: You understand governed retrieval.
4. What should be logged in a RAG trace?
Strong answer: Question, filters, source ids, scores, prompt version, answer status, citations, refusal reason, latency, and cost.
Red flag: "Only log the final answer."
Follow-up: Which logs are sensitive?
What this proves: You think about observability.
5. What is an execution pipeline?
Strong answer: The ordered runtime path: auth, cache, query rewrite, retrieval, rerank, prompt, safety, answer, logging.
Red flag: "It is just a script."
Follow-up: Which step can be asynchronous?
What this proves: You can design a shippable system.
6. Where do indexing and retrieval strategies fit?
Strong answer: Indexing is mostly offline, retrieval is online, and both shape answer quality.
Red flag: "There is only vector search."
Follow-up: When is hybrid search better?
What this proves: You connect offline and online RAG systems.
Production prep
The full source file includes 20 questions grouped by foundations, production judgment, evaluation, and architecture. Use the Markdown version for deeper written practice.