Budget end-to-end latency for a real-time voice agent and explain why time-to-first-audio differs from time-to-first-token.

This is a premium question from DeepOffer's 2026 ml system design interview bank. Members get the full model answer — what interviewers are really testing, the step-by-step reasoning, common mistakes to avoid — plus AI voice mock interviews with live follow-ups and a scored report.