Design a Multi-Agent Research System System Design Interview

1. Problem Statement

2. Architecture Discussion Map

Use this as one discussion aid, not a single correct answer. Your design should follow from the requirements, scale, and trade-offs you establish.

graph LR
    A["Design a Multi-Agent Research System"]
    A --> F1["Workflow versus agent decision"]
    A --> F2["Planner/orchestrator and subagent contracts"]
    A --> F3["Context isolation, tool access, and result provenance"]
    A --> F4["Budgeting, cancellation, retries, and partial failure"]
    A --> F5["Trace-based evaluation and quality measurement"]

3. Key Focus Areas

• 1
  Workflow versus agent decision
• 2
  Planner/orchestrator and subagent contracts
• 3
  Context isolation, tool access, and result provenance
• 4
  Budgeting, cancellation, retries, and partial failure
• 5
  Trace-based evaluation and quality measurement

4. What Strong Candidates Should Demonstrate

• ✓Choose multi-agent architecture only when parallelism justifies its cost.
• ✓Design context isolation, delegation, synthesis, and citation provenance.
• ✓Evaluate non-deterministic workflows using traces and task-level graders.

Evaluation Guide

This is an evaluation framework, not a single model answer. Strong designs may make different choices when their assumptions and trade-offs are explicit.