Microsoft Foundry: Build AI Apps and Agents (AI-103) is a hands-on Python course that covers the skills measured in the AI-103 exam, from secure deployment and RAG with Azure AI Search to agents with tools, MCP and multi-agent workflows, evaluation, guardrails and vision, speech and document solutions. It suits Python developers, AI engineers and anyone preparing for the AI-103 exam.
Building a chatbot prototype on a language model takes very little time, but taking AI apps and agents into real use across an organisation is another matter. Development teams have to choose the right models and services for each task, bring in organisational data so the AI answers correctly, let agents call tools safely, control costs and access, measure answer quality and deal with risks such as prompt injection, all on a platform that can be maintained over the long term.
This course covers the skills measured in the AI-103 Developing AI Apps and Agents on Azure exam, following the latest outline on Microsoft Learn, with learners writing Python code in Microsoft Foundry in their own Azure subscription. Day one covers planning and managing solutions: infrastructure, model deployment, security, cost and CI/CD. Day two builds RAG with Azure AI Search and agents that use
tools, MCP and multi-agent workflows with human approval steps. Day three measures quality with evaluation and tracing, sets up guardrails and then moves on to image, video, text, speech and document extraction with Content Understanding, before a closing capstone. The course helps with exam preparation, but learners book the exam with Microsoft themselves. (3 days, 6 hours per day, 18 hours in total, Intermediate level.)
Course Details
This course covers the skills measured in the AI-103 Developing AI Apps and Agents on Azure exam, following the latest outline on Microsoft Learn. It runs for 3 days, 6 hours per day (18 hours in total, 09:00-16:00), as lectures with Python coding labs. Intermediate level. It builds on our RAG & Knowledge Base and other agent courses, focusing on building and running solutions on Microsoft Foundry. Every lab runs in the learner's own Azure subscription (a free account, pay-as-you-go or a company sandbox);
model tokens, Azure AI Search, Content Understanding, Speech and image or video generation models are billed by actual usage at the learner's own cost, and each lab ends with clean-up steps. The course helps with exam preparation but is not official Microsoft courseware; it does not include the exam fee or a voucher, and learners book the exam with Microsoft themselves. Learners take home a lab guide, Python code for every lab, sample CI/CD files, an evaluation question set and a summary of the skills in the exam outline.
Section 1: Lab: Microsoft Foundry and the AI-103 Outline
Section 2: Lab: Choosing and Deploying Models
Section 3: Lab: Security and Connecting Applications
Section 4: Lab: Quotas, Costs and CI/CD
Section 5: Lab: Prompts and Tuning Model Behaviour
Section 6: Lab: Retrieval and Indexing with Azure AI Search
Section 7: Lab: Building a RAG Application
Section 8: Lab: Building Agents with Foundry Agent Service
Section 9: Lab: Tools, MCP and Knowledge
Section 10: Lab: Multi-agent Solutions and Human Approval
Section 11: Lab: Evaluation, Tracing and Monitoring
Section 12: Lab: Responsible AI and Guardrails
Section 13: Lab: Generating and Editing Images and Video
Section 14: Lab: Multimodal Understanding
Section 15: Lab: Text Analysis, Speech and Document Extraction
Section 16: Workshop: Capstone and Exam Preparation
No public rounds are open right now. Join the waiting list and we will contact you first when the next round opens, or ask us on LINE. Or call 02-570-8449 or 088-807-9770
RAG, or retrieval-augmented generation, first searches an organisation's documents for relevant content and then passes it to the model to answer from, so responses are grounded in real, current data. Azure AI Search supports several retrieval methods: keyword search for exact terms, vector search for meaning via embeddings, hybrid search combining the two, and the semantic ranker, which reorders results by relevance.
MCP, the Model Context Protocol, is an open standard for connecting AI agents to external tools and data sources. Each system exposes its capabilities through an MCP server in a common format, so an agent can discover and call those tools without a custom connector for every system. Combined with function calling in Foundry Agent Service, it lets an agent read data or act on other systems within the permissions you set.
An API key is a secret stored in code or configuration; if it leaks, anyone can use it, and it has to be rotated regularly. A managed identity lets an app running on Azure authenticate through Microsoft Entra ID without storing any secret, with RBAC granting access only to the resources it needs. This keyless approach reduces the risk of leaks and makes access easier to audit.
It suits Python developers building AI apps and agents on Azure, AI engineers and solution developers, platform and DevOps teams running AI systems, and people preparing for the AI-103 exam. It is intermediate level: you should write Python fluently, including calling REST APIs with JSON, understand prompts, tokens and embeddings at AI-901 level, and know Azure basics such as resource groups, storage and Microsoft Entra ID.
You will be able to choose and deploy models on Microsoft Foundry secured with managed identity and RBAC, build RAG with Azure AI Search, create agents using function calling, MCP and multi-agent workflows, measure quality with evaluation, tracing and guardrails, and extract data from documents with Content Understanding. Participants receive a Certificate of Completion with its own number that can be verified online, separate from Microsoft's AI-103 exam.
Bring a laptop with VS Code, Python, the Azure CLI and Git, plus your own Azure subscription. Usage of models, Azure AI Search, Content Understanding, Speech and image generation is billed to you at actual consumption, with steps to delete resources after each lab. The course helps you prepare but is not official Microsoft training, does not include an exam voucher, and you register for the exam yourself.
AI is now part of almost every team's work, from chatbots that answer customers and document summaries to meeting transcripts and automatic invoice reading. Yet many people are still unsure which kind of AI suits each task, how the models actually work and…
General AI models are strong at language but do not know your organization's internal information - manuals, policies, contracts, customer data or accumulated know-how. When asked something specific to the organization, AI often answers incorrectly…
AI today is more than a chatbot that answers questions. Agentic AI can plan and carry out work for us, built around two open standards: MCP (Model Context Protocol), which gives AI a single standard way to use tools, files and databases, and A2A…
