Your question is Near-Duplicate Video Detection System. 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).
Design a near-duplicate / visually similar video detection system for a very large short-video platform, covering large-scale embedding retrieval, embedding compression/quantization for high query volume, handling false positives via human review and appeals, and quantitatively modeling the tradeoff between embedding dimensionality, retrieval quality, and serving latency. Asked in the ML / system design stage. Candidate said this was the hardest round; felt weakest on the human-review calibration and quantitative quality-vs-latency questions. Reported follow-ups: Scale to production-size corpus; Human-in-the-loop appeal process