AI-103 Azure AI apps and agents developer
Design, build and run AI applications and agents on Microsoft Foundry, with .NET: services and deployment, agents and tools, images and video, text and speech, information extraction. The topics of exam AI-103, building on the AI track.
Search the academy Study plan for this course
Who it is for
For .NET developers who have to build and run AI applications on Azure or who are aiming at exam AI-103 and already know the AI track.
Prerequisites
The AI track (at least modules T1 to T7 and T10) or equivalent experience with Microsoft.Extensions.AI, RAG, agents and evaluation, plus K2 and the Azure basics from K3. You need an Azure account only for the optional steps; model availability and quotas differ per subscription and region. The exam itself expects Python experience: the course teaches .NET and only reads Python.
Afterward, you can
- Choose Foundry services, models and deployment types from requirements and justify the choice
- Design infrastructure, security, quotas, monitoring and responsible AI for a Foundry solution
- Build agents with the Foundry SDK: tools, memory, several agents, approvals, evaluation
- Generate images and videos, understand images and analyze content with Content Understanding
- Analyze and translate text, process speech and use it as an agent's modality
- Ingest, enrich and index content and connect it to agents
Contents
- I0
3 h
Orientation
map of the exam and the AI track, the exercise account with cost and clean-up, reading Python and writing .NET
- I1
8 h
Plan and manage
choosing services and models, infrastructure and CI/CD, quotas and monitoring, security, responsible AI in operation
- I2
12 h
Generative and agentic solutions
Foundry SDK, RAG, agents and tools, several agents and approvals, evaluation, tuning, observability
- I3
7 h
Images and video
generating and editing images and videos, understanding images, Content Understanding, responsible AI for images
- I4
6 h
Text and speech
text analysis, translation, speech to text and back, speech as an agent's modality
- I5
7 h
Information extraction
indexing, semantic and hybrid search, enrichment, Content Understanding for documents, retrieval for agents
- I6
4 h
Wrap-up
final project: the travel assistant as a design and sketch, defending the design, checking gaps
About 47 hours in all (indicative).
Lessons
You see the text and video of every lesson here on the site after buying the course; the titles and goals are already shown. A self-study course: you work through the lessons in order and check your exercises yourself against the solutions.
You can read the first module (I0) for free, with text and video, without signing in: so you see how the course works before you buy.
The lessons stand under their modules, in the order of the course. Click a module to open or close its lessons.
I0 Orientation
Indicative: 3 hours · 3 lessonsfree to read
Lesson 1 · about 60 minutes
The map: exam, AI track and this course
Lesson 2 · about 60 minutes
The practice account: Foundry resource, project, costs, cleanup
Lesson 3 · about 60 minutes
Reading Python, writing .NET: bringing the exam's examples to C#
I1 Plan and manage
Indicative: 8 hours · 5 lessons
Lesson 4 · about 90 minutes
Choosing Foundry services and models: decisions from requirements
Lesson 5 · about 90 minutes
Infrastructure, deployment types and CI/CD for Foundry projects
Lesson 6 · about 90 minutes
Quotas, scaling, cost and monitoring
Lesson 7 · about 90 minutes
Security: identity, keyless access, private network, roles
Lesson 8 · about 120 minutes
Responsible AI in operation: guardrails, evaluation, auditability, oversight
I2 Generative and agentic solutions
Indicative: 12 hours · 8 lessons
Lesson 9 · about 90 minutes
The Foundry SDK in .NET: connecting application and project, using models
Lesson 10 · about 90 minutes
RAG in the application with Foundry and Azure AI Search
Lesson 11 · about 90 minutes
Building agents: role, goal, conversation, memory, versions
Lesson 12 · about 90 minutes
Connecting tools: functions, APIs, MCP, search, Content Understanding, toolboxes
Lesson 13 · about 120 minutes
Multiple agents, workflows and approvals
Lesson 14 · about 90 minutes
Evaluating models and agents, analyzing errors
Lesson 15 · about 90 minutes
Tuning behavior: prompts, parameters, reflection, self-critique
Lesson 16 · about 60 minutes
Observability, multiple models and combining rules
I3 Images and video
Indicative: 7 hours · 5 lessons
Lesson 17 · about 90 minutes
Generating and editing images from text and references
Lesson 18 · about 90 minutes
Generating and editing videos
Lesson 19 · about 90 minutes
Understanding images: captions, questions about images, alt text
Lesson 20 · about 90 minutes
Content Understanding for image and video; objects and regions
Lesson 21 · about 60 minutes
Responsible AI for images: filters, hidden instructions, visual rules
I4 Text and speech
Indicative: 6 hours · 4 lessons
Lesson 22 · about 120 minutes
Text analysis with language models and Foundry Tools
Lesson 23 · about 60 minutes
Translating: Azure Translator or a language model
Lesson 24 · about 90 minutes
Speech to text and text to speech, custom speech models
Lesson 25 · about 90 minutes
Speech as an agent modality, understanding audio, translating speech
I5 Information extraction
Indicative: 7 hours · 5 lessons
Lesson 26 · about 90 minutes
Ingesting and indexing content: indexers, data sources, media
Lesson 27 · about 90 minutes
Semantic, hybrid and vector search for grounding
Lesson 28 · about 90 minutes
Enrichment with skills: OCR, layout, custom skills
Lesson 29 · about 90 minutes
Content Understanding for documents: analyzers, fields, Markdown
Lesson 30 · about 60 minutes
Connecting retrieval pipelines to workflows and agent tools
I6 Wrap-up
Indicative: 4 hours · 2 lessons
Lesson 31 · about 150 minutes
Capstone project: the travel assistant as a design and sketch
Lesson 32 · about 90 minutes
Defending the design, checking gaps, next steps
Final project: the Kladde
What the Kladde can do after this course
No further work on the Kladde but a final project: the travel assistant of the made-up Nordlicht Reisen GmbH as a decision folder with an architecture overview and a Bicep sketch, compiled and not deployed.
At the end you have designed every area of the exam on one continuous case and built or sketched it in .NET, and assembled a decision folder. You check the final project yourself against a list.
Certifications
An offer, never a requirement: these Microsoft exams fit this stage. You take them at Microsoft; we are not a Microsoft training partner.
AI-103 Azure AI Apps and Agents Developer Associate
Expects Python skills.
The course follows the topics Microsoft lists for exam AI-103 (the version of the skills list dated April 16, 2026, read on Microsoft Learn on November 12, 2026) and replaces neither Microsoft's material nor exam preparation. The exam expects Python experience; the course teaches .NET. You take the exam with Microsoft; registration, price and dates are there.
Keep learning with Microsoft Learn
Microsoft's learning paths and documentation that go with this course.