Your question is In-Memory Vector Embedding Cache. 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 an in-memory cache specifically optimized for storing and retrieving high-dimensional vector embeddings? Discuss the cache API, vector memory layout, keying, eviction, sharding, replication, consistency, and persistence choices. Explain how you would meet latency and memory constraints, handle cache misses and embedding version changes, and validate performance under realistic load.
