AI architecture guidance to build AI workloads on Azure
This article offers architecture guidance for organizations running AI workloads on Azure. It focuses on Azure AI platform-as-a-service (PaaS) solutions, including Azure AI Studio, Azure OpenAI, Azure Machine Learning, and Azure AI Services. It covers both generative and nongenerative AI workloads.
The Azure Architecture Center offers reference architectures and guides to help organizations build AI workloads efficiently and securely. These resources provide well-tested, structured frameworks for AI workload deployment. In AI Ready, you established a resource hierarchy that categorizes AI workloads into internal and internet-facing groups. Deploy AI workloads to subscriptions under the appropriate management groups (internal vs. internet-facing). The following tables list articles for building AI workloads.
Note
If you're using Azure landing zones, begin with the Baseline Azure OpenAI architecture in Azure landing zone and deploy it to an application landing zone subscription.
Generative AI architectures and guides
| Article | Article type | Target organization |
|---|---|---|
| Baseline Azure OpenAI architecture in an Azure landing zone | Architecture | Enterprise |
| Baseline Azure OpenAI reference architecture | Architecture | Any |
| Basic Azure OpenAI reference architecture | Architecture | Startup |
| GenAIOps | Guide | Any |
| Developing RAG solutions | Guides | Any |
| Proxy Azure OpenAI usage | Guide | Any |
Nongenerative AI architectures and guides
| Article | Article type | Target organization |
|---|---|---|
| Document processing architectures | Architectures | Any |
| Video and image classification architecture | Architectures | Any |
| Audio processing architecture | Architecture | Any |
| Predictive analytics architecture | Architecture | Any |
| Azure Machine Learning | Guides | Any |
| MLOps | Guides | Any |
| Team Data Science Process | Guides | Any |
Use the AI design areas as a framework
The AI design areas provide technology-specific framework to design AI workloads with Azure's AI platform-as-a-service (PaaS) solutions. It focuses on Azure AI Studio, Azure OpenAI, Azure Machine Learning, and Azure AI Services. Use them to establish standards and best practices related to these services:
Each design area includes recommendations for both generative and nongenerative AI workloads on Azure, consolidating best practices that apply to all AI workloads using Azure PaaS AI platforms.