179 questions
No questions match those filters.
← All questions
LLMOps & ProductionHard· Embedding Models· Index Maintenance· Compute Costs
The retrieval team wants to ship a new embedding model...
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 plansThe disagreement is a consequence of the efficiency trade-off: precomputed vectors are bound to the specific model weights. Re-encoding 500M chunks is computationally expensive and requires significant storage for dual-write windows during cutover. If a shadow evaluation shows the new model provides minimal recall gains, the platform team is correct to block it. If the gains are substantial, build a blue/green index path to automate the process.