Skip to main content
  1. Public practice questions
  2. AZ-104 Practice Exams | Azure Administrator Associate
  3. AZ-104 Case Study Set

Question 14 — AZ-104 Case Study Set

Paideffort Adaptive AI Platform Deployment
 

Business Requirements

Paideffort is standardizing Azure deployments for an AI-assisted certification platform used by development, test, and production teams. The same infrastructure definition must support different storage SKUs, service capacities, regional settings, and naming conventions without requiring engineers to edit the ARM template before each deployment. Development should normally use Standard_LRS storage, while production administrators must be able to select Standard_GRS or Standard_ZRS when resilience requirements change.

Security and governance requirements

Application authentication must use Microsoft Entra ID wherever supported. Deployment credentials, passwords, and secrets must not remain readable from deployment history. Production pipelines must consume deployment results without repeatedly querying Azure for values already produced during provisioning. Infrastructure changes must remain traceable through version-controlled templates and environment-specific parameter files.

Application Architecture

The application uses a Microsoft Foundry project endpoint for agent and project operations and an Azure OpenAI endpoint for direct OpenAI-compatible model operations. DefaultAzureCredential provides Microsoft Entra authentication. A FunctionTool performs deterministic eligibility calculations, an OpenAPI tool retrieves certification catalog information, and Azure Functions process longer-running document-generation jobs. The platform must eventually support additional agents, higher throughput, regional expansion, and centralized observability.

from azure.identity import DefaultAzureCredential
from azure.ai.projects import AIProjectClient

credential = DefaultAzureCredential()

project_client = AIProjectClient(
    endpoint=FOUNDRY_PROJECT_ENDPOINT,
    credential=credential
)

agent_settings = {
    "function_tool": "eligibility_check",
    "openapi_tool": "course_catalog",
    "async_processor": "certificate-document-function"
}

ARM Template Design

The infrastructure team created one ARM template for all environments. The template is intended to let release engineers select an environment-specific storage SKU and provide an application administrator secret during deployment. Storage names follow Paideffort's internal naming convention, while deployment outputs are consumed by later pipeline stages.

{
  "$schema": "https://schema.management.azure.com/schemas/2019-04-01/deploymentTemplate.json#",
  "contentVersion": "3.0.0.0",
  "parameters": {
    "environmentName": {
      "type": "string",
      "allowedValues": [ "dev", "test", "prod" ],
      "metadata": {
        "description": "Deployment environment"
      }
    },
    "storageAccountType": {
      "type": "string",
      "defaultValue": "Standard_LRS",
      "allowedValues": [
        "Standard_LRS",
        "Standard_GRS",
        "Standard_ZRS"
      ],
      "metadata": {
        "description": "Storage account SKU"
      }
    },
    "adminSecret": {
      "type": "string",
      "defaultValue": "Paideffort-Temporary-2026"
    }
  },
  "variables": {
    "storageName": "[concat('pdf', parameters('environmentName'), uniqueString(resourceGroup().id))]"
  },
  "resources": [
    {
      "type": "Microsoft.Storage/storageAccounts",
      "apiVersion": "2025-01-01",
      "name": "[variables('storageName')]",
      "location": "[resourceGroup().location]",
      "sku": {
        "name": "[parameters('storageAccountType')]"
      },
      "kind": "StorageV2",
      "properties": {
        "supportsHttpsTrafficOnly": true
      }
    },
    {
      "type": "Microsoft.Web/serverfarms",
      "apiVersion": "2024-11-01",
      "name": "[concat('plan-', parameters('environmentName'))]",
      "location": "[resourceGroup().location]",
      "sku": {
        "name": "[parameters('functionPlanSku')]"
      }
    }
  ],
  "outputs": {
    "storageEndpoint": {
      "type": "object",
      "value": "[reference(variables('storageName')).primaryEndpoints]"
    },
    "storageAccountName": {
      "type": "string",
      "value": "[variables('storageName')]"
    }
  }
}

Deployment Process

Development engineers normally accept the template's default storage SKU. Production deployments use an environment parameter file. After deployment, the release workflow needs the storage account name and its runtime endpoint so that later application configuration stages can reference the deployed resource.

- name: Deploy production infrastructure
  run: |
    az deployment group create \
      --name paideffort-prod \
      --resource-group $RESOURCE_GROUP \
      --template-file ./infra/main.json \
      --parameters environmentName=prod storageAccountType=Standard_GRS

- name: Discover storage settings
  run: |
    STORAGE_NAME=$(az storage account list \
      --resource-group $RESOURCE_GROUP \
      --query "[0].name" -o tsv)

    STORAGE_ENDPOINT=$(az storage account show \
      --resource-group $RESOURCE_GROUP \
      --name $STORAGE_NAME \
      --query "primaryEndpoints.blob" -o tsv)
Future platform direction

Paideffort expects additional environments to use different capacity settings, tool endpoints, and storage configurations. The platform team wants parameter definitions to constrain invalid deployment choices, document administrator inputs, and protect confidential values. Deployment outputs must provide downstream automation with selected resource names and runtime properties. Repeated deployment of an unchanged template to the same environment must leave existing resources in the same desired state.