Separate total and active parameters, then size persistent state, approximate compute and token dispatch. Compare supported parallel layouts using per-rank memory, measured communication and recovery.
Updated Sep 2026 · Learn how AI infrastructure works through explanations, worked examples and diagrams, then practise applying it to interview questions.
Separate total and active parameters, then size persistent state, approximate compute and token dispatch. Compare supported parallel layouts using per-rank memory, measured communication and recovery.
The full answer is part of Premium. A free account includes more answers in each topic, but does not unlock this one.
Use these concept explanations to review the ideas behind this question.
Scored on separating total from active parameters in the first minute, on deriving the layout from the memory and communication budgets rather than naming a fashionable one, and on treating expert imbalance as a straggler problem with a measurable signal.
No comments yet — be the first to share your approach.
