Skip to content
Linear Horizon Training

Ready-made course · 2 days · Intermediate

AI Systems on Azure

A two-day course for cloud engineers, developers and architects: design enterprise AI systems on Azure, including identity, search, model APIs and the architecture those services should sit inside.

Azure is often chosen for enterprise AI because of identity, compliance and the Microsoft estate already in place. That is a reason to be more careful about architecture, not less. This course teaches how to design AI systems on Azure without treating a managed agent or a search resource as the whole design.

The course covers model APIs, Microsoft Entra integration, Azure AI services, RAG and search, storage, APIs, agents, security, observability, networking, deployment and cost. Identity is a thread through the two days: who the user is, what they may retrieve, and what a tool may do.

As with the AWS course, services are implementations. You design the system first, then map it onto current Azure AI, search and hosting capabilities.

You leave able to build an enterprise-shaped assistant or application on Azure with identity, retrieval and operations that would survive a security review.

Who it is for
Cloud engineers, developers and architects working on Azure who will design or implement enterprise AI systems.
Before you start
Working knowledge of Azure identity, resource organisation and a compute option. Familiarity with Entra ID and with LLM applications is helpful.

You will be able to

  • Design an Azure AI architecture with Entra identity, data boundaries and a model-access path.
  • Implement model APIs and Azure AI services behind an application contract, not as the application itself.
  • Build RAG using Azure storage and search with permissions that match enterprise identity.
  • Integrate agents and tools with network, identity and API constraints.
  • Apply security, observability, networking and cost controls appropriate to an enterprise estate.
  • Deploy a working system and review it as a security and operations artefact.

Modules

8 quarter-day modules. Each one can go into a custom course on its own.

  1. AI services on Azure

    • Azure AI Foundry, Azure OpenAI and the wider AI services.
    • Subscriptions, regions, deployments and quotas.
    • Reference architecture for an AI application on Azure.
    • Cost model: pay-as-you-go and provisioned throughput.
  2. Azure OpenAI in practice

    • Deployments, model versions and API versions.
    • Chat, structured outputs and tool calling.
    • The Foundry model catalogue beyond OpenAI.
    • Lab: call and compare deployed models.
  3. RAG on Azure

    • Azure AI Search: indexes, vectors and semantic ranking.
    • Indexers, skillsets and document parsing.
    • Security trimming for per-user results.
    • Lab: build an index and a grounded chat over it.
  4. Agents on Azure

    • Foundry Agent Service.
    • Tools, connections and Logic Apps.
    • Semantic Kernel and other frameworks.
    • Lab: deploy an agent with tools.
  5. AI APIs on Azure

    • API Management as an AI gateway.
    • Functions and Container Apps for AI services.
    • Token limits, load balancing and semantic caching.
    • Lab: put a gateway in front of two deployments.
  6. Identity, networking and content safety on Azure

    • Entra ID and managed identities instead of keys.
    • Private endpoints and network isolation.
    • Azure AI Content Safety and prompt shields.
    • Data residency and abuse monitoring settings.
  7. Operating AI on Azure

    • Monitoring and tracing in Foundry and Application Insights.
    • Evaluations in Foundry.
    • Quotas, throttling and capacity planning.
    • Cost management and chargeback.
  8. Lab: an AI application on Azure, end to end

    • Put the pieces together: API, model, search index and agent.
    • Deploy with Bicep or Terraform.
    • Add content safety, tracing and an evaluation run.
    • Review the design against the Azure Well-Architected guidance.

Practical work

You build an Azure-backed path: Entra-authenticated access, retrieval over a small corpus, and an application or agent that calls a constrained API. The review focuses on identity, network, cost and operations.

You take away

  • A working Azure implementation of a small enterprise AI path
  • Reference architectures mapped to current Azure AI and search services
  • Identity, network and cost checklists
  • A mapping from portable architecture to Azure resources

Combine it with

Contact