GCP Generative AI Leader Practice Question 72: ML Lifecycle Deployment Stage
Google Cloud Generative AI Leader · Test 8 · question 2 of 10
A data science team has finished training and evaluating a custom classification model and has selected the version that meets the accuracy bar. The team's director now wants to move that specific model version into a production environment where it can serve low-latency online predictions to a customer-facing application, and to be able to track which version is live at any time.
Which stage of the ML lifecycle is the director asking about, and what Google Cloud capability supports it?
Show answer and explanation
Answer:
Model deployment is the ML lifecycle stage where a trained, evaluated model version is made available to serve predictions in production. On Google Cloud, that means registering the model version in the Model Registry and deploying it to an endpoint, which provisions the serving infrastructure and gives the team a stable interface plus visibility into which version is currently live.
B is wrong because data ingestion happens before training, when raw data is first collected and loaded. C is wrong because data preparation is also a pre-training stage focused on cleaning, transforming, and labeling data, not serving a finished model. D is wrong because model training is the stage where the model learns from data; the scenario describes a model that has already been trained and evaluated and now needs to be served, which is deployment, not training.
Reference: Deploy a model to an endpoint