Exam-focused study notes for Microsoft Exam AI-103: Developing AI Apps and Agents on Azure, the exam behind the Microsoft Certified: Azure AI Apps and Agents Developer Associate certification.
📖 Read them at marcogrimaldi29.com/ai-103-study-notes
Written against the skills measured as of April 16, 2026.
| Page | Covers | Exam weight |
|---|---|---|
| Foundry Foundations | The Foundry resource and project model, Foundry Models, deployment types, Foundry Tools, connections, the Responses API, RBAC — and the full classic-to-new naming translation table | Prerequisite |
| Skill 1 · Plan and manage | Choosing models and services, infrastructure design, CI/CD, quotas and rate limits, monitoring and drift, keyless security, responsible AI guardrails | 25–30% |
| Skill 2 · Generative AI and agents | RAG, prompt and hosted agents, tool schemas and function calling, toolboxes, multi-agent orchestration, approvals, evaluation, prompt engineering, observability | 30–35% |
| Skill 3 · Computer vision | Image and video generation, inpainting and masks, multimodal understanding, accessible alt text, Content Understanding standard and pro mode, visual responsible AI | 10–15% |
| Skill 4 · Text analysis and speech | Azure Language vs generative prompting, structured JSON output, sentiment and safety, translation, the speech stack and Voice Live | 10–15% |
| Skill 5 · Information extraction | Azure AI Search indexers and skillsets, integrated vectorization, vector/hybrid/semantic search, agentic retrieval, Content Understanding analyzers | 10–15% |
| SDK deep dive | The Python the exam makes you read: AIProjectClient, the Responses API, the function-calling loop, agents, evaluation, tracing, plus a classic-to-new migration map |
— |
| Exam tips & caveats | Key numbers, decision trees, rebrand traps, confused service pairs, a scenario-to-answer lookup table and a pre-exam checklist | Final review |
Every page carries Mermaid diagrams, decision tables, Python samples and exam-caveat callouts.
- Light / dark mode with a switcher drawn as the site's own sunrise brand mark, which opens into a full sun in light mode
- Two-rail navigation — a foldable left sidebar for moving between pages (collapsing to an icon rail, then to a drawer), and a right-hand On this page outline with a marker that slides between sections as you scroll. Below the aside breakpoint the outline folds into a collapsible card between the page title and the body
- Mermaid diagrams that re-render with the active theme
- Build-time syntax highlighting with Shiki — no highlighter ships to the browser
- Floating Home and Back to top buttons that appear on scroll
- Accessible by design: skip link, visible focus, semantic landmarks, reduced-motion support, print stylesheet
- Cookieless analytics via Umami
src/
├── components/ Header, Footer, Sidebar, OnThisPage, ThemeToggle, Mermaid, CodeBlock, Callout…
├── data/site.ts Single source of truth: site identity + page registry
├── layouts/ BaseLayout (head, header, footer) and NoteLayout (sidebar + content)
├── lib/urls.ts Base-path helpers for links and assets
├── pages/ One directory per note page
└── styles/ global.css — the whole design system
public/images/ Brand mark and profile image
Navigation, sidebar, footer and pager all read from the page registry in src/data/site.ts.
Astro · Mermaid · Shiki · Umami — all open source.
These notes were researched and written with the help of AI — Claude in this case — working from the official Microsoft Learn documentation, the AI-103 study guide and the AI-103T00-A course syllabus. Every page was reviewed before publication, but AI-assisted writing can still get details wrong, and Microsoft Foundry changes quickly. Treat these notes as a study companion and verify anything decision-critical against the official documentation.
Foundry changes fast and Microsoft renames features often, so if something here is wrong or out of date, please report an issue.
These notes are an independent study aid for learning purposes only. They are not affiliated with or endorsed by Microsoft. Always verify against the official AI-103 study guide and the Microsoft Foundry documentation before relying on any detail here.
If these notes helped you prepare for AI-103, there are a couple of ways to show it: ⭐ star this repo on GitHub so other candidates can find it, or 🤝 connect with me on LinkedIn.
Maintained by Marco Grimaldi, Cloud Solution Architect. Find more study notes and certification reviews at marcogrimaldi29.com, or reach out through my contact page.