Identify Generative AI Services and Capabilities in Microsoft Azure AI-900 Practice Questions (Page 9)

Questions 41–45

1. 41application · medium

   A team is evaluating whether to use Azure AI Foundry for their generative AI project. They need to support multiple models, compare outputs, and deploy to production with monitoring. Which capability of Azure AI Foundry directly supports this requirement?

   • AAutomated machine learning
   • BUnified model catalog and deployment endpoints
   • CData labeling and annotation
   • DCognitive search integration
   Reveal Solution Discussion Select an answer first
2. 42foundation · easy

   What capability of Azure AI Foundry directly supports the deployment of generative AI models to production?

   • AIntegration with Azure Machine Learning pipelines
   • BOne-click deployment to managed endpoints
   • CAutomated model training
   • DAutomatic scaling of all models without configuration
   Reveal Solution Discussion Select an answer first
4. 43expert · hard

   A company wants to use a specific open-weight model from the Azure AI Foundry model catalog. The model is available in multiple versions. The company needs to ensure that their application always uses the same version, even after new versions are released. They also need to be able to roll back quickly if a problem occurs. What should they do?

   • ACopy the model to a custom container and host it on Azure Container Instances
   • BUse Azure OpenAI Service to deploy the same model
   • CUse the latest version of the model in the catalog
   • DDeploy the specific model version to a managed endpoint and pin the version in the deployment configuration
   Reveal Solution Discussion Select an answer first
5. 44application · medium

   A team is evaluating models in the Azure AI Foundry model catalog. They need to compare models based on inference cost, latency, and quality metrics. What is the most efficient way to make this comparison?

   • AContact each model provider for custom quotes
   • BUse the model catalog's filtering and comparison features
   • CTrain a custom model to mimic all candidates
   • DDeploy each model to a separate endpoint and test manually
   Reveal Solution Discussion Select an answer first
6. 45expert · hard

   A company is building a generative AI application that must handle sensitive customer data. They require that all data stays within a specific Azure region and that network traffic to the model endpoint does not traverse the public internet. They also need to monitor token usage. Which combination of Azure OpenAI Service features should they implement?

   • ADeploy in the required region, use a private endpoint, and enable diagnostic logging
   • BEnable fine-tuning and use Azure AD authentication
   • CUse Azure AI Foundry prompt flow and content filtering
   • DUse the model catalog and deploy to a public endpoint
   Reveal Solution Discussion Select an answer first