Sample question 1 — Collaborating Within and Across Teams
A platform engineering team hardening a customer-facing system has this requirement: a biased recommendation caused customer complaints and needs a clear escalation path. Which specific action listed below is the right choice given the constraints?
- A. Define model incident response playbooks with rollback, communication, and root-cause
- B. Orchestrate feature materialization as upstream pipeline steps with dependencies.
- C. Publish reusable pipeline components in a shared repository for validation and splitting.
- D. Enable audit logging for prediction access and restrict endpoint IAM.
Answer: A. Option A is correct: "Define model incident response playbooks with rollback, communication, and root-cause steps." This is the Google Cloud-aligned way to address incident response for model harm, which matches the scenario. Option B is incorrect because feature pipeline orchestration is the wrong control here — it is built for a team that offline feature computation must finish before training starts, not this scenario's constraint. Option C is incorrect because component reuse across teams addresses a team that every team reimplements the same data validation and split components, not the failure mode this stem describes for incident response for model harm. Option D is incorrect because Selecting audit logs for prediction access would apply "Enable audit logging for prediction access and restrict endpoint IAM", which fits a different production need than the one above.