Your Goal: Learn how to design agentic AI systems and structure your answer to interview questions like "Design an AI customer support agent" or "Design ChatGPT".
Session Overview
- About 20 minutes with Coach Arjun, who walks you through 14 slides shown on your screen
- Topics: the agent loop, workflow vs agent, reference architecture, async workers and streaming, context management, tools and permissions, guardrails, cost and latency, single vs multi-agent, evals
- Short questions along the way and one multiple-choice check
- A final practice round: you design an AI customer support agent out loud
How to Get the Most Out of This
- Answer in your own words; your reasoning matters more than the exact term
- Say "I don't know" if stuck; the coach will give a hint
- Ask to go back to a slide any time
- In the practice round, drive the design yourself like in a real interview
What Scores High
- Explaining why, not just naming components
- Knowing when NOT to use an agent
- Covering stop conditions, idempotency, permissions and guardrails in your design
- Using feedback from the coach in your later answers
What Loses Points
- One-word answers
- Jumping to multi-agent or the biggest model by default
- Skipping safety for irreversible actions like refunds