Microsoft Azure AI Apps and Agents Developer (AI-103) Exam Prep

What's inside

1. Level 1: Exam Orientation and Exam Craft for AI-103Free
   [Exam orientation, supporting all sections] Explains what AI-103 measures under the skills measured as of April 16, 2026, and the five weighted sections that shape this course. It also covers how to read scenario 'best answer', product-contrast and recognition questions, and states the ground rules: the study guide is Microsoft's, is free on Microsoft Learn and is not reproduced here; this course is independent of Microsoft and not endorsed by it; and no pass is promised.
2. Level 2: Choosing Models: LLMs, SLMs, Multimodal and Foundry ToolsFree
   [Plan and manage an Azure AI solution – Choose the appropriate Foundry services for generative AI and agents] Shows how to choose an appropriate model for each task among large language models, small language models, multimodal models and prebuilt Foundry Tools, using the Foundry model catalog, benchmarks and the trade-offs between cost, latency and capability.
3. Level 3: Choosing Foundry Services for Each Workload
   [Plan and manage an Azure AI solution – Choose the appropriate Foundry services for generative AI and agents] Maps workload types (generative tasks, grounding, vector search, agent workflows and multimodal processing) to the Foundry service that fits each one, such as Foundry Models, Foundry Agent Service, Azure AI Search and Content Understanding.
4. Level 4: Choosing Retrieval and Indexing Methods
   [Plan and manage an Azure AI solution – Choose the appropriate Foundry services for generative AI and agents] Covers choosing an appropriate method for retrieval and indexing: keyword, vector, hybrid and semantic ranking, chunking and embedding choices, and when to index content versus query a source live.
5. Level 5: Choosing Memory, Tool and Knowledge Services for Agents
   [Plan and manage an Azure AI solution – Choose the appropriate Foundry services for generative AI and agents] Teaches how to choose appropriate memory, tool and knowledge integration services for agent solutions: conversation state, built-in knowledge tools, Azure AI Search, function calling, OpenAPI tools and MCP servers.
6. Level 6: Designing Azure Infrastructure for AI Apps and Agents
   [Plan and manage an Azure AI solution – Set up AI solutions in Foundry] Covers designing Azure infrastructure for AI apps and agent-based solutions: Foundry resources and projects, connected resources such as storage, Azure AI Search and Key Vault, region selection, and how to set up the environment through the portal, the Azure CLI or Bicep.
7. Level 7: Choosing Deployment Options and Configuring Deployments
   [Plan and manage an Azure AI solution – Set up AI solutions in Foundry] Explains how to choose appropriate deployment options (global, data zone or regional; standard versus provisioned throughput; Foundry-hosted models versus managed compute) and how to configure model and agent deployments, including versions, capacity and upgrade policy.
8. Level 8: Integrating Foundry Projects with CI/CD Pipelines
   [Plan and manage an Azure AI solution – Set up AI solutions in Foundry] Covers integrating Foundry projects with continuous integration and continuous deployment (CI/CD) pipelines: infrastructure-as-code, promotion across environments, pipeline identities, and evaluation gates in GitHub Actions or Azure Pipelines.
9. Level 9: Quotas, Rate Limits, Scaling and Cost
   [Plan and manage an Azure AI solution – Manage, monitor, and secure AI systems] Teaches how to manage quotas, scaling, rate limits and cost footprints for model and agent workloads: tokens-per-minute and requests-per-minute limits, handling HTTP 429 responses with retry and backoff, provisioned versus pay-as-you-go capacity, and cost tracking.
10. Level 10: Monitoring Model Performance, Drift, Safety and Grounding
    [Plan and manage an Azure AI solution – Manage, monitor, and secure AI systems] Covers monitoring model performance, drift, safety events and grounding quality in production, using Foundry monitoring, Azure Monitor and Application Insights, and continuous evaluation.
11. Level 11: Monitoring Ingestion Quality, Index Health and Relevance
    [Plan and manage an Azure AI solution – Manage, monitor, and secure AI systems] Covers monitoring data ingestion quality, search index health and relevance performance in Azure AI Search: indexer execution history and errors, document counts and storage, and measuring retrieval relevance over time.
12. Level 12: Securing AI Systems: Identity, Keyless Auth and Networking
    [Plan and manage an Azure AI solution – Manage, monitor, and secure AI systems] Covers configuring security, including managed identity, private networking, keyless credentials and role policies: Microsoft Entra ID with DefaultAzureCredential, Azure RBAC roles for Foundry, private endpoints and disabling public network access.
13. Level 13: Safety Filters, Guardrails and Content Moderation
    [Plan and manage an Azure AI solution – Implement responsible AI across generative AI and agentic systems] Teaches how to configure safety filters, guardrails, risk detection and content moderation: content filter categories and severity thresholds, prompt shields for direct and indirect attacks, blocklists, and Azure AI Content Safety.
14. Level 14: Responsible AI Instrumentation: Evaluators and Safety Evals
    [Plan and manage an Azure AI solution – Implement responsible AI across generative AI and agentic systems] Covers applying responsible AI instrumentation, including evaluators, safety evaluations and explanation tooling: built-in risk and safety evaluators, adversarial simulation and red-teaming runs, and reading the explanations evaluators give for their scores.
15. Level 15: Auditing and Governing Agent Behaviour
    [Plan and manage an Azure AI solution – Implement responsible AI across generative AI and agentic systems] Covers implementing auditing through trace logging, provenance metadata and approval workflows, and governing agent behaviour with oversight modes, constraints and tool-access controls.
16. Level 16: Connecting an App to a Foundry Project with the SDKs
    [Implement generative AI and agentic solutions – Build generative applications by using Foundry] Teaches how to configure an application to connect to a Foundry project and how to integrate generative workflows by using the Foundry SDKs and connectors: the project endpoint, the Foundry project client in Python and C#, connections, and getting an OpenAI-compatible client.
17. Level 17: Deploying and Consuming LLMs, Small, Code and Multimodal Models
    [Implement generative AI and agentic solutions – Build generative applications by using Foundry] Covers deploying and consuming LLMs, small models, code models and multimodal models from code and REST: chat and Responses-style calls, streaming, image inputs, and structured outputs.
18. Level 18: Implementing RAG in an Application
    [Implement generative AI and agentic solutions – Build generative applications by using Foundry] Teaches how to implement retrieval-augmented generation (RAG) in an application: querying an Azure AI Search index, building grounded prompts, returning citations, and handling the case where nothing relevant is retrieved.
19. Level 19: Workflows, Tool-Augmented Flows and Multistep Reasoning
    [Implement generative AI and agentic solutions – Build generative applications by using Foundry] Covers designing workflows, tool-augmented flows and multistep reasoning pipelines: function calling from a model deployment, chaining steps, choosing reasoning models, and handling tool errors.
20. Level 20: Evaluating Models and Apps: Fabrication, Relevance, Quality
    [Implement generative AI and agentic solutions – Build generative applications by using Foundry] Teaches how to evaluate models and apps, including detecting fabrications, relevance, quality and safety: groundedness, relevance, coherence and fluency evaluators, test datasets, AI-assisted versus metric-based scoring, and comparing runs in the Foundry portal.
21. Level 21: Defining Agents: Roles, Goals and Tool Schemas
    [Implement generative AI and agentic solutions — Build agents by using Foundry] You learn how to define a Foundry Agent Service agent's role, goals and instructions, choose how it keeps track of a conversation, and write tool schemas that the model can call reliably. Examples use the Python and C# SDKs and the Foundry portal.
22. Level 22: Agents with Retrieval, Function Calling and Memory
    [Implement generative AI and agentic solutions — Build agents by using Foundry] You build an agent that grounds its answers on uploaded files with file search, runs your own code through function calling, and remembers context within a conversation and across conversations.
23. Level 23: Agent Tools: APIs, Search and Content Understanding
    [Implement generative AI and agentic solutions — Build agents by using Foundry] You connect agents to external APIs through OpenAPI tools, to knowledge stores and Azure AI Search indexes through project connections, to Content Understanding for document and media input, and to custom functions. You also learn which tool fits which exam scenario.
24. Level 24: Connecting Agents to MCP Servers
    [Implement generative AI and agentic solutions — Build agents by using Foundry] You connect a Foundry agent to a remote Model Context Protocol (MCP) server with the MCP tool: setting the server label and URL, limiting which tools it may call, setting up authentication, and deciding when MCP is a better choice than a custom function or OpenAPI tool.
25. Level 25: Orchestrating Multi-Agent Solutions
    [Implement generative AI and agentic solutions — Build agents by using Foundry] You build multi-agent solutions in Foundry, where specialised agents work together through workflow orchestration patterns (sequential, hand-off and group-style) and agent-to-agent (A2A) calls. You also learn that classic connected agents are deprecated in the new Foundry Agent Service.
26. Level 26: Semiautonomous Agents with Approval Flows
    [Implement generative AI and agentic solutions — Build agents by using Foundry] You build autonomous and semiautonomous agent workflows with safeguards: human approval before tool calls, MCP approval settings, step and token limits, and points where a person must sign off before the workflow continues.
27. Level 27: Monitoring, Evaluating and Debugging Deployed Agents
    [Implement generative AI and agentic solutions — Build agents by using Foundry] You add monitoring to deployed agents, score their behaviour with agent evaluators (intent resolution, tool-call accuracy, task adherence), and run error analysis on traces to find out why a run failed or went off track.
28. Level 28: Tuning Generation: Prompts and Model Parameters
    [Implement generative AI and agentic solutions — Optimize and operationalize generative AI systems] You tune how a model generates output by improving system messages, few-shot examples and output formats, and by adjusting parameters such as temperature, top_p, max tokens, stop sequences and reasoning effort. You also learn which parameters reasoning models don't accept.
29. Level 29: Reflection, Chain-of-Thought Checks and Self-Critique
    [Implement generative AI and agentic solutions — Optimize and operationalize generative AI systems] You build model reflection and self-critique loops (generate, critique, revise), evaluate reasoning quality with chain-of-thought evaluations, and cap these loops so they don't run up cost and latency.
30. Level 30: Observability: Tracing, Tokens, Safety and Latency
    [Implement generative AI and agentic solutions — Optimize and operationalize generative AI systems] You set up end-to-end observability with OpenTelemetry tracing into Application Insights, so you can analyse token usage, follow safety signals and break down latency across model calls, tool calls and retrieval.
31. Level 31: Orchestrating Multiple Models and Hybrid Rules
    [Implement generative AI and agentic solutions — Optimize and operationalize generative AI systems] You orchestrate several models and flows by routing each request to a small or large model as needed, chaining specialised models, and combining an LLM with a deterministic rules engine for decisions that must be repeatable and auditable.
32. Level 32: Generating Images from Text and Reference Media
    [Implement computer vision solutions — Design and implement image- and video-generation solutions] You build a solution that generates images from text prompts and reference images with Foundry image models such as GPT-image-1, through the SDK, REST and the Foundry portal playground.
33. Level 33: Image Editing: Inpainting, Masks and Controls
    [Implement computer vision solutions — Design and implement image- and video-generation solutions] You set up image-editing workflows, including inpainting, mask-based edits and prompt-driven changes, and you choose the generation and editing controls the platform offers, such as size, quality, number of images and output format.
34. Level 34: Generating and Editing Videos with Sora
    [Implement computer vision solutions — Design and implement image- and video-generation solutions] You build video-generation solutions with Sora models in Foundry, from text prompts and from image or video input. You handle the asynchronous job pattern, edit generated videos with features such as remix, and choose controls such as resolution and duration.
35. Level 35: Multimodal Visual Analysis and Captions
    [Implement computer vision solutions — Design and implement multimodal understanding workflows] You build solutions that analyse what is shown in images by sending them to multimodal models, and you set up the app to produce short or detailed captions for one image or for several.
36. Level 36: Visual Question Answering and Accessible Alt-Text
    [Implement computer vision solutions — Design and implement multimodal understanding workflows] You implement question answering that is grounded in what an image actually shows, and you generate alt-text and longer image descriptions that follow accessibility guidelines.
37. Level 37: Content Understanding for Images: Analyzers and Modes
    [Implement computer vision solutions — Design and implement multimodal understanding workflows] You set up Azure Content Understanding in Foundry Tools to extract visual characteristics and to identify objects, components or regions in images. You also build the single-task and pro-mode pipelines that the guide names, and check which API version supports each.
38. Level 38: Video Analysis Workflows with Content Understanding
    [Implement computer vision solutions — Design and implement multimodal understanding workflows] You build video analysis workflows that split a video into segments, describe and interpret each segment, and identify objects or regions over time, using Content Understanding video analyzers.
39. Level 39: Responsible AI for Visual Content
    [Implement computer vision solutions — Implement responsible AI for multimodal content] You set up filters that classify unsafe or disallowed images, detect and stop indirect prompt injection hidden as text inside images, and enforce visual policy rules such as watermarking and content credentials, flagging prohibited symbols, brand-usage requirements and inappropriate content.
40. Level 40: Extracting Entities, Topics and Structured JSON
    [Implement text analysis solutions — Apply language model text analysis] You extract entities, topics, summaries and structured JSON from text, both with generative prompting and structured outputs and with Foundry Tools language capabilities, and you choose between the two approaches.
41. Level 41: Speech to Text and Text to Speech for Agents
    [Implement text analysis solutions — Implement speech solutions] Learn how to convert speech to text and text to speech for agentic interactions, using the Azure Speech SDK in Python or C#, the REST endpoints, Speech playground steps in the Foundry portal, SSML voice control, and the Voice Live API as a single real-time speech-to-speech loop in front of a Foundry agent.
42. Level 42: Custom Speech and Multimodal Audio Reasoning
    [Implement text analysis solutions — Implement speech solutions] Learn to integrate speech as an agent modality, including custom speech models trained on domain vocabulary, and to enable multimodal reasoning from audio inputs with audio-capable and realtime models deployed in Foundry.
43. Level 43: Translating Speech with Foundry Tools and LLMs
    [Implement text analysis solutions — Implement speech solutions] Learn to translate speech into other languages with Azure Speech translation in Foundry Tools (speech to translated text and to synthesised speech) and with language-model translation flows such as LLM speech, and when to choose each.
44. Level 44: Ingesting and Indexing Multimodal Content
    [Implement information extraction solutions — Build retrieval and grounding pipelines] Learn to ingest and index documents, images, audio and video into Azure AI Search with data sources, indexers, index schemas and the portal's import wizard, including how non-text media becomes searchable text and vectors.
45. Level 45: Semantic, Hybrid and Vector Search for Grounding
    [Implement information extraction solutions — Build retrieval and grounding pipelines] Learn to configure vector search (HNSW or exhaustive KNN, vectorizers), hybrid search with Reciprocal Rank Fusion, and the semantic ranker, and to choose the right combination for grounding a generative app.
46. Level 46: Enrichment with Built-in and Custom Skills
    [Implement information extraction solutions — Build retrieval and grounding pipelines] Learn to build skillsets that enrich content with built-in skills for text, images and layout (OCR, Document Layout, splitting, embedding, entity extraction) and custom Web API skills, with output field mappings into the index.
47. Level 47: OCR-Based RAG Ingestion and Agent Retrieval Tools
    [Implement information extraction solutions — Build retrieval and grounding pipelines] Learn to configure a full RAG ingestion flow for scanned and born-digital documents using OCR and layout-aware chunking, and to connect the resulting retrieval pipeline directly to workflows and agent tools, such as the Azure AI Search tool and Foundry IQ knowledge bases with agentic retrieval.
48. Level 48: Multimodal Document Extraction: OCR, Layout, Fields
    [Implement information extraction solutions — Extract content from documents] Learn to extract information with multimodal pipelines that combine OCR, layout analysis and field extraction, using Document Intelligence prebuilt and custom models and Content Understanding, and to choose between the two.
49. Level 49: Content Understanding Analyzers for Agents and RAG
    [Implement information extraction solutions — Extract content from documents] Learn to implement Content Understanding analyzers that produce clean, grounded markdown or structured field output for agents and RAG, using prebuilt RAG analyzers such as prebuilt-documentSearch and custom analyzers with field schemas.
50. Level 50: Timed Mock Exam: 50 Questions in 100 Minutes Timed mock
    A timed mock exam of 50 original scenario questions in 100 minutes, drawn from all five AI-103 skill areas in proportion to the exam weightings, with every distractor explained and a 70% pass mark.

Access

The first 2 levels are free with a free account. Every level, the podcast edition and the AI tutor come with All Access at £4.99/month or any Creator plan — see pricing.

Start free →