What's changed: Added per-section figures (cert-figure-retrofit). New AI-901 Chapter 2 (Domain 2: Foundry overview = project/model catalog/playground/connected services/evaluation & monitoring/portal vs SDK; model selection criteria and deployment = endpoint + key, token billing, context length)
2.1Overview of Microsoft Foundry
Understand Microsoft Foundry, the unified platform for building and operating AI solutions—its portal, projects, model catalog, connected Azure AI services, playground, and evaluation/monitoring—plus how the portal and SDK divide the work.
Microsoft Foundry (formerly Azure AI Foundry, earlier Azure AI Studio) is the unified platform to build, evaluate, and operate AI solutions in one place. Work that used to be scattered—models here, services there, deployment on another screen—comes together behind one front door. You can: (1) pick a model from the catalog, (2) deploy it to get an endpoint, (3) try it in the playground, (4) assemble generative AI apps and agents, and (5) evaluate and monitor. AI-901 Domain 2 centers on implementing on Foundry.
2.1.1Projects — the unit of work
Foundry organizes work into projects. A project gathers your deployed models, connected Azure AI services, data, agents, and evaluation results, and can be shared with a team—so you manage "which model, with which settings, combined with which data" in one place. A project is backed by Azure resources, so billing and access control (roles) follow Azure’s mechanisms.
2.1.2Key parts of Foundry
- Model catalog (Foundry Models): where you choose from many pretrained models—Microsoft, OpenAI, open models, and more.
- Deployment: make a chosen model usable and obtain an endpoint and key.
- Playground: a no-code chat surface to try prompts and inspect behavior.
- Connected Azure AI services: connect Language, Speech, Vision, Content Understanding, etc. to the project to use alongside models.
- Evaluation & monitoring: measure and continuously monitor response quality, safety, and cost.
2.1.3Portal vs SDK
Foundry has two entry points. The Foundry portal (web UI) is best for point-and-click work—browsing/deploying models, trying prompts in the playground, and configuring evaluations. The Foundry SDK (mainly Python) is for calling deployed endpoints from code and embedding them in apps and agents. A useful rule: explore/configure in the portal, implement apps with the SDK. AI-901 does not test deep coding, but know the SDK’s role as "how an app calls a model."
Foundry does not replace the individual Azure AI services—it unifies them into one dev environment. If you need translation or sentiment analysis, you still use Azure AI Language and connect it to your Foundry project. Think of Foundry as a unified front door/workspace, not an all-in-one replacement service.
Watch "task → where you do it" questions. E.g., "deploy a model and get an endpoint," "try prompts with no code (playground)," "continuously measure quality/safety/cost (evaluation & monitoring)" all happen in Foundry. "Generic translation/sentiment API" = Azure AI Language (connected to Foundry). The portal = UI work, the SDK = calling from code.
2.1.4Section summary
- Microsoft Foundry = one platform for select → deploy → build app/agent → evaluate → monitor
- The unit of work is the project (gathers models/services/data/agents/evaluations; backed by Azure resources)
- Key parts: model catalog / deployment / playground / connected services / evaluation & monitoring
- Portal = UI/exploration; SDK = call models from code and embed in apps
Sign in to track progress — Log in.
Quick check
(just a quick review)Q1. Which unified platform lets you select, deploy, try (playground), evaluate, and monitor models in one place?
Q2. In Foundry, what unit of work gathers deployed models, connected services, data, and evaluation results?
Q3. Which Foundry feature lets you try prompts and inspect model behavior without writing code?
Q4. What do you use to call a deployed model from application code and embed it in an app?
Q5. Which correctly describes the relationship between Foundry and the individual Azure AI services?
Q6. Where in Foundry do you choose from many pretrained models (Microsoft, OpenAI, open models)?