Prepare and test your skills

Prepare and test your skills

Free sample question with answers and detailed explanation below. The Cloud Digital Leader practice bank has an additional 3,070 practice questions.Try 10 free realistic questions: instant score, no credit card, and no account required.
A retail organization wants to create a competitive advantage by building a custom machine learning model using its proprietary customer purchasing data. The data science team needs an end-to-end workflow in Vertex AI that meets the following requirements:
Which workflow using Vertex AI managed services should the team implement?
Vertex AI Vizier, Vertex AI TensorBoard, and Vertex AI Feature Store are specialized managed services within the Vertex AI ecosystem that support hyperparameter optimization, experiment visualization, and feature management, respectively.
These services would be the right choice if the team specifically needed to store and share reusable ML features in Vertex AI Feature Store, optimize hyperparameters with Vertex AI Vizier, and visually debug training performance metrics using Vertex AI TensorBoard as supplemental tools within their broader workflow.
Understanding the Vertex AI lifecycle requires distinguishing between features (the processed data inputs stored in Vertex AI Feature Store) and models (the trained mathematical constructs stored in Vertex AI Model Registry and deployed to endpoints).
Vertex AI provides a unified, end-to-end platform of managed services that guides data science teams through every stage of the custom machine learning (ML) development lifecycle—from initial data exploration and preparation to model training, hyperparameter tuning, evaluation, version registration, and production deployment.

Unlike fragmented or self-managed workflows that require extensive DevOps intervention at every transition point, combining Vertex AI Workbench, Vertex AI Experiments, Vertex AI Model Registry, and managed Vertex AI endpoints provides a cohesive, automated path from prototype to production.
A self-managed machine learning workflow relies on manual scaling of compute instances, tracking unmanaged artifacts outside of a centralized ML platform, and deploying trained models onto self-managed container orchestration platforms such as Google Kubernetes Engine (GKE).
This approach would be appropriate if the organization had a dedicated platform engineering team that required strict, fine-grained control over custom container orchestration on Google Kubernetes Engine (GKE) and wanted to manage its own infrastructure rather than using a fully managed ML environment.
When deploying models for online inference, organizations must choose between a fully managed environment (using Vertex AI endpoints where Google Cloud instantiates and manages the physical machine resources) and a self-managed environment (such as GKE). For teams seeking rapid experimentation and minimal operational overhead, fully managed services are the optimal choice.
Prompt design is the process of crafting instructions and examples to guide the output of pre-trained foundation models, while Vertex AI Extensions connect generative models to external APIs and real-time systems so they can perform actions or retrieve external data.
This approach would be appropriate if the organization wanted to build a generative AI application using an existing pre-trained foundation model out of the box—guiding its responses with instructions and connecting it to external tools via Vertex AI Extensions—rather than training or tuning a custom model.
When building AI solutions on Vertex AI, organizations typically progress from prompt design and grounding (using pre-trained foundation models) to model tuning or full custom training when they need deep customization and business differentiation using their own proprietary training sets.
Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated