Zentraler Informatikdienst

Using AI models via Microsoft Foundry

For:  students    employees

This user guide will help you to deploy and use AI models via the Microsoft Foundry in Azure.

Requirements

You have ordered Microsoft Azure and already have access to the Microsoft Azure portal and the services it includes.

Hinweis

Diese Anleitung basiert auf der englischsprachigen Oberfläche von Microsoft Azure. Um zu dieser Spracheinstellung zu wechseln:

  1. Klicken Sie im Azure Portal in der blauen Leiste oben auf das Zahnrad-Symbol ⚙.
  2. Wählen Sie im Menü links Sprache und Region.
  3. Wählen Sie im Drop-down Sprache den Eintrag English.

  1. Open Azure portal – Microsoft Foundry – Create a resource
  2. Enter Subscription, Resource group, Name, Region, Default project name – tab Review + create
  3. Open resource – Go to Foundry portal
  4. Model catalog – select model – Use this model
  5. Change Deployment type (must not contain Global) – Deploy
  6. Models + endpoints – Open in playground or Endpoint

Detailed user guide

1. Creating Foundry resource

  1. Open the Azure portal.
  2. Enter the term Microsoft Foundry in the search bar at the top.
  3. Select the Microsoft Foundry service under Services in the search results.
  4. In the Create a Foundry Resource box click Create a resource.

2. Configuring Foundry resource

  1. Select the Subscription and the Resource group in which the resource is to be created.
  2. Enter a Name for the Foundry resource.
  3. Select the Region, i.e. the Azure data centre in which the resource is to be created.
    Due to data protection and data security measures, only some European Azure regions are available within the University of Vienna’s Azure environment. Depending on the selected region, the available models may vary. If you wish to use the latest models, such as ChatGPT-5.2, the ZID currently recommends the Sweden Central region.
  4. Enter a Default project name.
  5. The other tabs (Storage, Network etc.) can be configured if you require a specific setup. Otherwise, click the last tab Review + create and create the resource.

3. Opening Foundry resource and portal

  1. After a few seconds, a notification will appear indicating that the resource has been created successfully.
  2. Open the newly created Foundry resource.
  3. Click Go to Foundry portal.

4. Deploying model

Note

The models are updated by Microsoft on a regular basis.

  1. The Foundry portal will open in a new browser tab.
  2. To deploy a model, click Model catalog in the menu on the left.
  1. Select a model to deploy by clicking its name. You can also use the search bar to discover available models.
  2. After you have selected the desired model, you can view further details on the specific model page.
  3. To deploy the model to your Foundry Project, click Use this model.

5. Configuring model

  1. A configuration menu will open.
  2. Change the Deployment type to one that does not contain Global in its name, such as Data Zone Standard. This ensures that data is processed only in the European Union. 
    Deployment types that do not guarantee data processing within the EU are blocked by policy and cannot be deployed.

Note

If you need to use a model that is only available with the deployment type Global Standard and you don’t intend to process any sensitive data: Contact digital.zid(at)univie.ac.at describing your use case. The use case will be sent to the Data Protection officer of the University of Vienna. 

Only after receiving approval can the blocking policy be deactivated for your Azure subscription.

  1. Click Deploy.
  2. Click Models + endpoints in the menu on the left to view the deployment. This view also contains a complete list of all your deployments in your Foundry project.

6. Using model

  1. Click Models + endpoints in the menu on the left.
  2. Click on the desired model.
  3. Click Open in playground to use the model in your active Browser session.
  4. Alternatively, you can retrieve the Endpoint and API Key for embedding in an external application.

To top