Sierra take-home agent interview

Round 3 of 6

See the whole loop
Format
Build a customer-support AI agent for a made-up client and choose which features to ship; an OpenAI key is provided
They grade
Product judgment, prioritization, agent fundamentals, metrics and observability, design rationale
AI use
The agent must call the model and its tools directly, without an agent framework
Heads-up
Treat it as production code; you defend and extend it onsite.

Questions

No Sierra reports yet. Other labs ask: