Your question is ML System Design for OpenAI Search/Recommendation. 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).
ML system design: search/recommendation system that adapts LLMs to search, ranking and retrieval.
Asked in the phone screen stage. This topic summary was quoted from the recruiter, and a commenter confirmed receiving this question.
Explain how you would clarify requirements, design the retrieval, ranking, and re-ranking pipeline, decide which components use LLMs, and evaluate and operate the system. Address online versus batch processing, latency and cost, feedback loops, training-serving skew, feature drift, and failure modes.