RAG & retrieval
Chunking, embeddings, hybrid search, reranking, citations, retrieval evals, freshness, permissions on customer data.
- Explain RAG to the customer's head of engineering. When would you pick it over fine-tuning for their private data? · easy · reported 2×Free
- Design a RAG system over a customer's 40,000 internal documents (policies, contracts, SOPs). Where does retrieval quality actually come from? · medium · reported 3×Free
- How do you chunk real enterprise content: contracts, a wiki page with a long pricing table, and support-ticket threads? · medium · reported 3×Free
- When would you use hybrid search instead of pure vector search, and how do you combine the results? · easy · reported 3×Free
- What does a reranker do, when is it worth the added latency, and what can it not fix? · medium · reported 2×Free
- The assistant must respect each user's document permissions. Why can't you just tell the model in the prompt, and how do you enforce it? · hard · reported 4×Pro
- An employee loses access to a folder, or deletes a document containing personal data. How do you make sure the assistant stops using it, and how fast? · hard · reported 3×Pro
- The source documents change every day. How do you design ingestion so the index stays correct without re-embedding everything? · medium · reported 2×Pro
- How do you make the citations in a RAG answer trustworthy, and what should the assistant do when the sources do not contain the answer? · easy · reported 2×Pro
- Retrieval returns the right documents but the assistant's answers are still wrong. What do you check? · medium · reported 2×Pro
- How do you choose an embedding model, and what happens when the customer wants to switch models six months in? · medium · reported 2×Pro
- What metadata do you attach to every chunk at ingestion, and what is each field for? · easyPro
- Users ask follow-ups like 'what about contractors?' in a chat. How do you handle multi-turn retrieval and keep the context window under control? · hard · reported 1×Pro
- The customer wants the RAG system inside their own cloud. How do you choose the vector store, and what goes wrong when you filter vector search by permissions or tenant? · medium · reported 2×Pro
- What are the security risks specific to RAG, and how do you protect a multi-tenant deployment? · hardPro