Design request shaping for long prompts, retrieval, and tool use.

Instruction: Describe how you would shape requests so complex workflows stay inside latency and cost budgets.

Context: Assesses whether the candidate can design a practical architecture and explain the main tradeoffs. Describe how you would shape requests so complex workflows stay inside latency and cost budgets.

Official answer available

Preview the opening of the answer, then unlock the full walkthrough.

I would shape the request before it hits the model: trim redundant context, keep only the retrieval results needed for the current subtask, summarize long history into durable state, and avoid sending tool outputs that do not influence the next decision.

I also like staged...

Upgrade to view official answer

Related Questions