Your question is Low-Latency RAG Pipeline Design. 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).
How would you design a RAG pipeline to ensure low-latency retrieval for real-time decision support at Anduril?
Explain how you would structure ingestion, indexing, retrieval, reranking, and generation for a system such as Lattice. Address latency and cost together, define how you would evaluate retrieval quality and groundedness offline and online, and describe defenses against hallucinations and prompt injection.
Your answer should identify practical engineering tradeoffs rather than assume unlimited infrastructure.