Your question is Design a Low Latency RAG Platform. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
You are building a retrieval augmented generation system over a very large document corpus. Users ask natural language questions and expect grounded answers with relevant citations from enterprise content.
How would you architect a RAG system that handles millions of documents with low latency?