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 financial services company has tuned a foundation model in Vertex AI using proprietary market data to create a differentiated customer advisory application. To operationalize this customized model into production, the company must meet the following requirements:
Which strategy should the company implement?
Prompt design is the practice of crafting natural language instructions and input-response examples to guide a foundation model's output without altering its underlying weights. Selecting the largest parameter model relies on a high parameter count to process complex patterns and relationships in data.
This approach would be the right choice if the company lacked a training dataset large enough for tuning, needed to handle highly complex, general-purpose reasoning where response quality outweighed latency and cost concerns, and could satisfy its requirements through prompt engineering alone.
Deploying models in a self-managed environment on Google Kubernetes Engine (GKE) involves packaging the model into containers and orchestrating the underlying Kubernetes nodes and GPU accelerators directly. Combining this with manual scaling means administrators explicitly adjust the number of compute instances or containers through direct intervention.
This option would be suitable if the organization required fine-grained control over custom container orchestration and infrastructure configurations on GKE, had a dedicated platform engineering team to manage clusters, and ran highly predictable batch workloads that only needed occasional manual capacity adjustments.
In Vertex AI, Google's standard, out-of-the-box foundation models come with pre-configured endpoints that are automatically available within a Google Cloud project. This allows developers to send API requests immediately to base models without explicitly provisioning machine types or deploying model artifacts.
This approach would be appropriate if the organization only needed a standard, untuned foundation model for general tasks—such as basic text summarization or drafting—and wanted the fastest possible go-to-market without customizing model weights or provisioning endpoint hardware.
Vertex AI Model Registry is a centralized repository that manages the lifecycle of machine learning models, organizing versions and enabling direct deployment to production. Deploying a custom or tuned model to a fully managed Vertex AI endpoint associates the model with dedicated physical machine and accelerator resources that Google Cloud automatically provisions, scales, and maintains for online, low-latency predictions.

Compared to self-managed container clusters or unconfigured endpoints, combining Vertex AI Model Registry with a fully managed Vertex AI endpoint directly supports end-to-end operational integration while balancing scalability, latency, and cost.
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