AI-900 Practice Question: Describe fundamental principles of machine learning on Azure

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✓

  Create an Azure Computer Vision resource, then obtain the endpoint and key, then call the API with the image and desired features, then parse the JSON response.

  Why this is correct

  This is the only correct order because Azure Computer Vision provides image analysis through an Azure resource that must first be provisioned in your subscription; only after creation does Azure generate the endpoint URL and subscription key that authenticate API requests. You then call the REST API, passing the image binary or URL, the desired visual features (e.g., Tags, Description, Categories), and the key in the Ocp-Apim-Subscription-Key header. Finally, the response is a JSON payload containing results like caption text, tag confidence values, and bounding boxes that your application needs to parse to extract meaningful insights. Without provisioning first, there are no credentials and no endpoint to call.

• ✗

  Obtain the endpoint and key, then call the API, then create the resource, then parse the response.

  Why it's wrong here

  This order is impossible because endpoint and key are dynamically generated only when you create a Computer Vision resource; you cannot obtain credentials for a resource that does not yet exist in your Azure subscription. Attempting to call the API before creating the resource would mean you have no endpoint URL and no subscription key to include in the request header, so the service would reject the request with an authentication or not-found error. Even if you somehow had credentials from another resource, those would not be scoped to the image-analysis task you intend to perform. Resource creation is the strict prerequisite that makes credential retrieval and API calls meaningful.

• ✗

  Call the API with the image, then parse the response, then create the resource, then obtain endpoint and key.

  Why it's wrong here

  This is incorrect because the Computer Vision API call itself is impossible without an Azure resource: without a valid endpoint and Ocp-Apim-Subscription-Key, your HTTP request will fail with 401 Unauthorized or 404 Not Found, so there is no successful JSON response to parse. The sequence also puts parsing before credential retrieval, which is illogical because parsing happens only after the API returns a response. Creating the resource and obtaining its endpoint/key must precede any call to the service. In short, you cannot 'call the API with the image' until you have been granted access through a provisioned resource.

• ✗

  Create the resource, then call the API, then obtain endpoint and key, then parse the response.

  Why it's wrong here

  This order fails because creating the resource alone does not automatically give you the endpoint and key in your code; you must explicitly retrieve them from the Azure portal or Azure CLI after provisioning and then embed them in the API request. Placing the API call before that retrieval means the call lacks authentication credentials, so the Computer Vision endpoint will reject it. The proper post-creation step is to obtain the credentials, not to jump directly into the call. Only after you have both the endpoint and key can you invoke the API and then parse the resulting JSON.

Go deeper

Related to this question