RAG system with document permissions
Design a retrieval-augmented-generation (RAG) serving system for an enterprise knowledge assistant over ~50M internal documents (wikis, tickets, code, PDFs). Users ask natural-language questions and get a grounded, cited answer in under 3 seconds p95. The corpus changes constantly (docs edited/deleted hourly), answers must respect per-user document permissions (a user must never see content from a doc they can't access), and answer quality and groundedness matter more than raw latency. Query volume is ~200 QPS.
What a strong answer looks like
Clarify scale and constraints first. Propose a clean component breakdown, then go deep on the hard parts (data model, bottlenecks, consistency, failure modes) and name the trade-offs you are making.
Clarify5:00 left
Estimate5:00 planned
Design15:00 planned
Deep dive12:00 planned
Failure8:00 planned
Which questions mattered is sealed until you submit. Telling you now would just be handing over the edge cases.
Run or narrate your approach, then ask the coach.