Skip to content

Azure AI Services Guide

🧠 Problem You need to build intelligent applications using prebuilt AI capabilities such as vision, language, speech, document understanding, or search. How do you provision and consume Azure AI services in a secure, manageable, and efficient way?

🔧 Azure Solution Use Azure AI services to consume cloud-based AI capabilities via REST APIs or SDKs. Provision services either as a multi-service or single-service resource depending on usage, scalability, and management needs.

📦 Required Components * Azure Subscription * Azure AI resource (multi-service or single-service) * Endpoint URI * Subscription Key(s) * Azure Region (Location) * REST API or SDK for preferred language (C#, Python, JavaScript, Java, Go)

🏗️ Architecture / Development

🛠 Provisioning Options: * Multi-service Resource: * One resource for multiple AI capabilities (Language, Vision, Speech, etc.) * Single billing, credentials, and endpoint * Single-service Resource: * Separate resources per service * Per-region customization * Independent billing and access management * Free tier often available

🧠 Training vs Prediction: * Some services (e.g., custom models) require separate training and prediction resources * Allows cost separation between model training and inference

🔑 Consuming Services: * Endpoint URI: HTTP address for service requests * Subscription Key: Authenticates client application * Location: Azure region for the resource * Two keys are provided; both are valid and regenerable

🧪 Using REST APIs: * Send data in JSON via HTTP (POST, PUT, GET depending on the function) * Receive JSON responses * Language-agnostic: works with cURL, Postman, or any programming language supporting HTTP

💻 Using SDKs: * Easier abstraction over REST * Available for major languages: * C# (.NET Core) * Python * JavaScript (Node.js) * Java * Go * Include service-specific classes and methods

✅ Best Practices / Considerations * Use multi-service resource for unified management in production * Use single-service resource when: * You need service separation (e.g., region, security, cost) * You're testing via the free tier * Regenerate keys regularly to enhance security * Use SDKs for robust application development; use REST for quick tests or language-agnostic integration * Always confirm service availability and SDK support per language and region in official documentation

❓Sample Exam Questions

  1. 🔐 You need to build an app that uses Azure AI Vision and Azure AI Speech. What provisioning option enables you to use a single set of credentials and billing? A. Separate single-service resources B. Multi-service resource ✅ C. Separate Azure subscriptions D. Cognitive Containers

  2. 🌍 Why might you choose to provision separate AI service resources for each service? A. To get better REST API performance B. To reduce the number of access keys C. To use different geographical regions or separate billing ✅ D. To increase SDK compatibility

  3. 🧾 Which information must be provided to a client application to consume an Azure AI service? (Choose all that apply) ⬜ Username and password ✅ Endpoint URI ✅ Subscription Key ✅ Resource Location (for some SDKs)

  4. 🧰 Which method allows language-agnostic access to Azure AI services? A. SDK B. Azure CLI C. REST API ✅ D. PowerShell

  5. 🧠 You want to build a custom model and keep training and prediction costs separate. What should you do? A. Use multi-service resource B. Use free-tier resources C. Use separate resources for training and prediction ✅ D. Enable autoscaling