Question 29
Domain 2 — AI Operations, Lifecycle, and Control EnvironmentA financial services firm operates at Google Cloud MLOps Maturity Level 1 with automated training pipelines and continuous training capabilities. The organization manually promotes models from staging to production after email-based approval from the model owner. Deployments occur during scheduled maintenance windows twice monthly. Monitoring detects performance degradation but requires manual intervention to roll back models. Which MLOps maturity advancement would most directly address the deployment bottleneck? (Select one!)
Correct answer: C
Explanation
Google Cloud MLOps maturity advances from manual promotion to automated deployment through CI/CD. The bottleneck is the manual staging-to-production step and rollback, so "CI/CD pipeline automation" with "approval gates" and "automated rollback capabilities" directly removes those delays while preserving control. This matches the need to move beyond email-based approvals and maintenance-window releases.
Why each option is right or wrong
A. Increase manual deployment frequency from twice monthly to weekly to reduce deployment latency
B. Add additional monitoring metrics to detect performance degradation earlier in the deployment cycle
C. Implement CI/CD pipeline automation enabling automated model deployment with approval gates and automated rollback capabilities
Google Cloud MLOps maturity Level 1 still relies on manual release steps, so the bottleneck here is the human-driven staging-to-production promotion and the twice-monthly maintenance-window constraint. Advancing to CI/CD automation with approval gates aligns with the standard deployment control pattern in MLOps and directly removes the manual email approval step; adding automated rollback addresses the failure-response gap when monitoring detects degradation, instead of waiting for manual intervention.
D. Implement feature flags allowing gradual traffic shifting between model versions during production deployment