AI-103 Azure AI Apps and Agents Course: Hands-On Guide

AI-103 Developing AI Apps and Agents on Azure Hands-On is a free video course that prepares you for the Microsoft AI-103 exam by building AI applications and agents with Microsoft Foundry and Python. It is for developers who want to go beyond a single chat call and learn the patterns behind real solutions: retrieval, agents with tools and memory, evaluation, security and deployment.

Each module is a hands-on Python lab with a shared toolkit, so you see code run and results printed on screen rather than slides alone. The full course is on the AI-103 course page.

Who this course is for

AI-103 is an Associate level exam: Developing AI Apps and Agents on Azure. The course suits:

• Python developers moving into generative AI and agent development on Azure.
• Azure developers who need to add Foundry models, retrieval and agents to their applications.
• AI engineers preparing for the AI-103 exam who want each skill shown as working code.

You need to be comfortable with Python. In the environment set-up lesson, the labs are described as needing Python 3.10 or later, with local simulators standing in for live Azure services.

What you will learn

The 25 videos group into six themes.

1. Set-up, Foundry and the exam

01. Environment Setup, Microsoft Foundry Portal Overview, 02. Foundry and the AI-103 Exam and 03. Python Development for Foundry build your working environment and map the exam’s skill domains onto the course. A reliable environment saves hours of confusion later, because you can tell a code problem from a set-up problem.

2. Models and generative AI apps

04. Model Selection and Deployment, 05. Building Generative AI Applications and 06. Reasoning Pipelines and Orchestration cover picking a model by capability, context length and cost, writing system prompts and structured outputs, and adding a critique-and-revise step to a pipeline.

3. Retrieval and grounding

07. Retrieval-Augmented Generation and 08. Azure AI Search for Grounding show how to stop a model inventing facts by retrieving the right text first. You see how chunk size changes retrieval results and why citations matter.

4. Agents

09. Foundry Agent Service, 10. Agent Tools and Function Calling, 11. Agent Memory and State, 12. Multi-Agent Solutions and 13. Autonomy with Safeguards build up an agent step by step: tools, short-term and long-term memory, several agents working together and limits on autonomous behaviour.

5. Quality, vision, speech and documents

14. Evaluation and Observability measures groundedness, relevance, coherence and safety. Then 15. Image and Video Generation, 16. Multimodal Understanding, 17. Responsible AI for Visual Content, 18. Text Analysis with Language Models, 19. Speech Solutions for Agents and 20. Document Extraction widen the course beyond text chat.

6. Security, deployment and the capstone

21. Security and Governance, 22. Deployment, CI/CD, Monitoring, Cost and 23. Capstone and Exam Preparation finish with what production needs: secure access, a deployment pipeline, monitoring and cost control, and a final review for the exam.

How the course is organised

• 25 videos, about 7.1 hours in total, free on YouTube.
• An 8-minute course overview, a short Foundry portal tour, then 23 numbered modules (01 to 23).
• Almost every module runs 16 to 20 minutes and follows the same rhythm: the problem, the idea, a lab script, its output and a recap.
• Modules build on each other, so the agent and retrieval work in later modules reuses ideas from earlier ones.

You can watch them in order in the YouTube playlist.

How to study it step by step

1. Watch the course overview, then the Foundry portal tour.
2. Get your environment working in Module 01. Check your Python version and run the first scripts before moving on.
3. Use Module 02 as your study map. Note which modules belong to which exam domain.
4. Learn the app basics (03 to 06). Rewrite at least one system prompt and see how the output changes.
5. Spend extra time on retrieval (07, 08). Retrieval-augmented generation sits underneath much of what follows.
6. Build the agent in order (09 to 13). Tools, memory, multiple agents and safeguards depend on each other.
7. Cover quality and the other modalities (14 to 20) one module per session.
8. Finish with security, deployment and the capstone (21 to 23).
9. Test yourself. Skills Pragati has a mock exam for AI-103 across five areas: planning and securing Foundry solutions, generative AI apps and retrieval, agents, vision, speech and documents, and evaluating, deploying and operating. Start with the free 10-question sample; the full timed mock comes with a subscription at ₹1,199/month or ₹7,999/year.

How it helps with the AI-103 exam

The course is built for AI-103: Developing AI Apps and Agents on Azure, at Associate level. Module 02 connects the exam’s skill domains to the labs, and Module 23 is a capstone with exam preparation. In between, every major skill area gets its own lab: Foundry and Python development, model choice, generative AI apps, RAG and Azure AI Search, Foundry Agent Service, tools, memory, multi-agent solutions, evaluation, vision, speech, document extraction, security and deployment.

Use the mock exam’s five areas to spot which modules to rewatch. Always check Microsoft’s official exam page for the current skills outline.

If you are new to Azure itself, the AZ-104 Azure Administrator course covers the platform fundamentals: identities, governance, networking and compute.

Frequently asked questions

Is the AI-103 course free?

Yes. All 25 videos are free on YouTube and listed on the AI-103 course page. Only the full mock exam needs a Skills Pragati subscription; the 10-question sample is free.

Do I need an Azure subscription for the labs?

According to the environment set-up lesson, the labs need Python 3.10 or later and use local simulators in place of live Azure services, so you can follow the patterns without provisioning or billing.

How long does AI-103 take?

The videos total about 7.1 hours. Running the scripts yourself and taking notes, three or four modules a week gets you through in six to eight weeks.

Is there a mock exam for AI-103?

Yes. It covers five areas, from planning and securing Foundry solutions to evaluating, deploying and operating them. Anyone can take the free 10-question sample.