Plan, Create and Deploy a Microsoft Foundry Service AI-102 Practice Questions (Page 6)

Questions 26–30

1. 26foundation · easy

   Which Azure service is commonly used to define and run CI/CD pipelines for deploying AI solutions?

   • AAzure Monitor
   • BAzure Pipelines
   • CAzure API Management
   • DAzure Logic Apps
   Reveal Solution Discussion Select an answer first
2. 27expert · hard

   A multinational company is deploying an AI model in Azure AI Foundry. They have a requirement that all traffic to the model must stay within the company's virtual network and not traverse the public internet. They also need to access the model from multiple on-premises sites. What should they do?

   • AUse the public endpoint and restrict access with IP whitelisting.
   • BUse Azure Front Door to route traffic to the model endpoint.
   • CConfigure a private endpoint for the Azure AI resource and connect it to the virtual network, then use a VPN gateway or ExpressRoute for on-premises connectivity.
   • DDeploy the model to Azure Kubernetes Service and use a load balancer with a public IP.
   Reveal Solution Discussion Select an answer first
4. 28expert · hard

   A company is deploying a model that must be available 24/7 but has very low traffic during off-peak hours. They want to minimize cost while maintaining availability. The model is containerized. Which deployment option should they choose?

   • ADeploy to Azure Container Instances (ACI) and start/stop it manually based on traffic.
   • BDeploy to Azure Machine Learning managed online endpoints with autoscaling and a minimum of one instance.
   • CDeploy to Azure Kubernetes Service (AKS) with a node pool that scales to zero during off-peak hours.
   • DDeploy to Azure Functions with the Consumption plan and rely on cold start handling.
   Reveal Solution Discussion Select an answer first
5. 29application · medium

   A logistics company has a custom object-detection model that must run on-premises at their warehouses due to intermittent internet connectivity. They want to use the same containerized model that they deploy in Azure, and they need to update the model periodically. What should they do?

   • AUse Azure Machine Learning to create a local web service on each warehouse server.
   • BPackage the model as a Docker container and deploy it to Azure IoT Edge devices at each warehouse.
   • CDeploy the model to Azure Container Instances and use a VPN connection from the warehouses.
   • DDeploy the model to Azure Kubernetes Service and access it over the internet from the warehouses.
   Reveal Solution Discussion Select an answer first
6. 30application · medium

   A healthcare startup is building a patient-facing symptom checker using Azure AI Foundry. They need to select a model that provides explainable reasoning for regulatory review. The team is evaluating a large language model (LLM) and a smaller, fine-tuned model. Which approach best aligns with Responsible AI principles and the solution's requirements?

   • AChoose the LLM because it is more accurate and document the model's general behavior in a policy manual.
   • BDeploy the LLM and rely on its built-in safety filters to ensure fairness and transparency.
   • CUse the smaller fine-tuned model and implement a separate explanation service to log reasoning traces for each prediction.
   • DSelect the smaller model but disable logging to reduce storage costs, as the model is less complex.
   Reveal Solution Discussion Select an answer first