Google Cloud Professional ML Engineer practice questions.

A data analyst with strong SQL but limited Python experience needs to train a model directly against a 5TB BigQuery dataset without moving data to a separate ML environment. Which approach fits?

A team needs to A/B test two model versions in production, routing 90% of traffic to the current model and 10% to a candidate. Which Vertex AI feature is the canonical fit?

A production model's prediction quality is drifting. The team needs to detect when input feature distributions shift significantly from the training distribution. Which Vertex AI capability is the canonical fit?

A team needs to define a reproducible ML pipeline (data prep, train, evaluate, register, deploy) that runs on Vertex AI infrastructure with versioning and lineage tracking. Which approach is the canonical fit?

A team prototyped a model in a Jupyter notebook on Vertex AI Workbench. They need to scale to a multi-GPU training job with hyperparameter tuning across 50 trials. Which approach is the canonical fit?