179 questions
No questions match those filters.
← All questions
System DesignHard· Retrieval· Latency· Data Processing
Our corpus is 50 GB of PDFs and we want sub-second answ...
This is one of the questions in the full AI/ML interview bank. Pro unlocks all 1789 questions; Premium includes the same bank plus the highest daily Practice limit.
See plansAt 50 GB, you are looking at approximately 42 million chunks, which exceeds the capacity for a single HNSW node at FP32. IVF-PQ is a more viable candidate for memory constraints. However, the real design risk is the generative decode time, which will consume most of the sub-second budget, making answer length and model size critical factors.