AI-103: Developing AI Apps and Agents on Azure · 5 skill areas · 64 skills
AI-103 study guide: every exam topic
Every topic on the AI-103 exam: Microsoft's 5 skill areas with their weights and all 64 skills, word for word, each with the episode of the free course that builds it and practice questions for it.
The skills are Microsoft's "Skills measured as of April 16, 2026", from Microsoft's AI-103 study guide. Microsoft updates its exams; check its guide for the current list.
Skills at a glance
Plan and manage an Azure AI solution 25–30%
Choose the appropriate Foundry services for generative AI and agents
-
Choose an appropriate model for each task, including large language models (LLMs), small language models, multimodal models, and Foundry Tools
Built in Episode 2: Choosing the model and the Foundry service for the job (coming)
-
Choose the appropriate Foundry services for generative tasks, grounding, vector search, agent workflows, or multimodal processing
Built in Episode 2: Choosing the model and the Foundry service for the job (coming)
-
Choose an appropriate method for retrieval and indexing
Built in Episode 2: Choosing the model and the Foundry service for the job (coming)
-
Choose appropriate memory, tool, and knowledge integration services for agent solutions
Built in Episode 2: Choosing the model and the Foundry service for the job (coming)
Set up AI solutions in Foundry
-
Design Azure infrastructure for AI apps and agent-based solutions
Built in Episode 1: AI-103 Tutorial, Ep 1: Microsoft Foundry Setup and Your First Model Call
-
Choose appropriate deployment options
Built in Episode 1: AI-103 Tutorial, Ep 1: Microsoft Foundry Setup and Your First Model Call
-
Configure model and agent deployments
Built in Episode 1: AI-103 Tutorial, Ep 1: Microsoft Foundry Setup and Your First Model Call
-
Integrate Foundry projects with continuous integration and continuous deployment (CI/CD) pipelines
Built in Episode 3: Managed identity, keyless credentials, quotas and cost (coming)
Manage, monitor, and secure AI systems
-
Manage quotas, scaling, rate limits, and cost footprints for model and agent workloads
Built in Episode 3: Managed identity, keyless credentials, quotas and cost (coming)
-
Monitor model performance, drift, safety events, and grounding quality
Built in Episode 3: Managed identity, keyless credentials, quotas and cost (coming)
-
Monitor data ingestion quality, search index health, and relevance performance
Built in Episode 3: Managed identity, keyless credentials, quotas and cost (coming)
-
Configure security, including managed identity, private networking, keyless credentials, and role policies
Built in Episode 3: Managed identity, keyless credentials, quotas and cost (coming)
Implement responsible AI across generative AI and agentic systems
-
Configure safety filters, guardrails, risk detection, and content moderation
Built in Episode 4: Content filters, guardrails and audit (coming)
-
Apply responsible AI instrumentation, including evaluators, safety evaluations, and explanation tooling
Built in Episode 4: Content filters, guardrails and audit (coming); Episode 7: Evaluate it: fabrications, relevance, quality, safety (coming)
-
Implement auditing through trace logging, provenance metadata, and approval workflows
Built in Episode 4: Content filters, guardrails and audit (coming)
-
Govern agent behavior with oversight modes, constraints, and tool-access controls
Built in Episode 4: Content filters, guardrails and audit (coming); Episode 10: Multi-agent orchestration and approval flows (coming)
Implement generative AI and agentic solutions 30–35%
Build generative applications by using Foundry
-
Deploy and consume LLMs, small models, code models, and multimodal models
Built in Episode 5: Deploy and call models from code with the Foundry SDK (coming)
-
Implement retrieval-augmented generation (RAG) in an application
Built in Episode 6: RAG on Azure: an app that answers from your documents (coming)
-
Design workflows, tool-augmented flows, and multistep reasoning pipelines
Built in Episode 6: RAG on Azure: an app that answers from your documents (coming)
-
Evaluate models and apps, including detecting fabrications, relevance, quality, and safety
Built in Episode 7: Evaluate it: fabrications, relevance, quality, safety (coming)
-
Integrate generative workflows into applications by using Foundry SDKs and connectors
Built in Episode 5: Deploy and call models from code with the Foundry SDK (coming)
-
Configure an application to connect to a Foundry project
Built in Episode 1: AI-103 Tutorial, Ep 1: Microsoft Foundry Setup and Your First Model Call
Build agents by using Foundry
-
Define agent roles, goals, conversation-tracking approach, and tool schemas
Built in Episode 8: Your first agent: role, tools, memory (coming)
-
Build agents that integrate retrieval, function-calling, and conversation memory
Built in Episode 8: Your first agent: role, tools, memory (coming)
-
Integrate agent tools, including APIs, knowledge stores, search, content understanding, and custom functions
Built in Episode 9: Agent tools: APIs, knowledge, search, custom functions (coming)
-
Implement orchestrated multi-agent solutions
Built in Episode 10: Multi-agent orchestration and approval flows (coming)
-
Build autonomous or semiautonomous workflows with safeguards and approval flow controls
Built in Episode 10: Multi-agent orchestration and approval flows (coming)
-
Integrate monitoring into deployed agents, evaluate agent behavior, and perform error analysis
Built in Episode 11: Tracing, token analytics and tuning (coming)
Optimize and operationalize generative AI systems
-
Tune generation behavior, such as prompt engineering and adjusting model parameters
Built in Episode 11: Tracing, token analytics and tuning (coming)
-
Implement model reflection, chain-of-thought evaluations, and self-critique loops
Built in Episode 11: Tracing, token analytics and tuning (coming)
-
Set up observability by implementing tracing, token analytics, safety signals, and latency breakdowns
Built in Episode 11: Tracing, token analytics and tuning (coming)
-
Orchestrate multiple models, flows, or hybrid LLM and rules engines
Built in Episode 11: Tracing, token analytics and tuning (coming)
Implement computer vision solutions 10–15%
Design and implement image- and video-generation solutions
-
Implement a solution that generates images from text prompts and reference media
Built in Episode 12: Generate and edit images and video (coming)
-
Implement a solution that generates videos from text prompts and reference media
Built in Episode 12: Generate and edit images and video (coming)
-
Configure image-editing workflows, including inpainting, mask‑based edits, and prompt‑driven modifications
Built in Episode 12: Generate and edit images and video (coming)
-
Implement workflows to edit generated videos
Built in Episode 12: Generate and edit images and video (coming)
-
Select and apply appropriate generation and editing controls provided by the platform
Built in Episode 12: Generate and edit images and video (coming)
Design and implement multimodal understanding workflows
-
Build a solution that analyzes visual context by using multimodal models
Built in Episode 13: Read images and video: captions, visual questions, Content Understanding (coming)
-
Configure apps to produce concise or detailed captions for single or multiple images
Built in Episode 13: Read images and video: captions, visual questions, Content Understanding (coming)
-
Implement a solution that enables question‑answering grounded in visual evidence
Built in Episode 13: Read images and video: captions, visual questions, Content Understanding (coming)
-
Configure generation of alt‑text and extended image descriptions aligned to accessibility guidelines
Built in Episode 13: Read images and video: captions, visual questions, Content Understanding (coming)
-
Implement visual understanding by configuring Azure Content Understanding in Foundry Tools to extract visual characteristics
Built in Episode 13: Read images and video: captions, visual questions, Content Understanding (coming)
-
Implement video analysis workflows to process and interpret video segments
Built in Episode 13: Read images and video: captions, visual questions, Content Understanding (coming)
-
Configure single‑task and pro‑mode Content Understanding pipelines
Built in Episode 13: Read images and video: captions, visual questions, Content Understanding (coming)
-
Implement solutions that identify objects, components, or regions within images or video
Built in Episode 13: Read images and video: captions, visual questions, Content Understanding (coming)
Implement responsible AI for multimodal content
-
Implement filters to classify unsafe or disallowed visual content
Built in Episode 13: Read images and video: captions, visual questions, Content Understanding (coming)
-
Detect and mitigate indirect prompt injection by using embedded text in images
Built in Episode 13: Read images and video: captions, visual questions, Content Understanding (coming)
-
Enforce visual policy rules, such as applying watermarks, flagging prohibited symbols, upholding brand usage requirements, and detecting potentially inappropriate content
Built in Episode 13: Read images and video: captions, visual questions, Content Understanding (coming)
Implement text analysis solutions 10–15%
Apply language model text analysis
-
Implement solutions to extract entities, topics, summaries, and structured JSON outputs by using generative prompting and Foundry Tools
Built in Episode 14: Entities, summaries, structured JSON, sentiment, translation (coming)
-
Configure detection of sentiment, tone, safety issues, and sensitive content
Built in Episode 14: Entities, summaries, structured JSON, sentiment, translation (coming)
-
Build solutions that translate text by using Azure Translator in Foundry Tools or LLM‑powered translation flows
Built in Episode 14: Entities, summaries, structured JSON, sentiment, translation (coming)
-
Customize language model outputs for domain tasks, such as compliance summarization and domain extraction
Built in Episode 14: Entities, summaries, structured JSON, sentiment, translation (coming)
Implement speech solutions
-
Implement workflows to convert speech to text and text to speech for agentic interactions
Built in Episode 15: Speech to text, text to speech, speech as an agent's input (coming)
-
Integrate speech as an agent modality, including custom speech models
Built in Episode 15: Speech to text, text to speech, speech as an agent's input (coming)
-
Enable multimodal reasoning from audio inputs
Built in Episode 15: Speech to text, text to speech, speech as an agent's input (coming)
-
Translate speech into other languages by using language models and Foundry Tools
Built in Episode 15: Speech to text, text to speech, speech as an agent's input (coming)
Implement information extraction solutions 10–15%
Build retrieval and grounding pipelines
-
Ingest and index content, such as documents, images, audio, and video
Built in Episode 16: Index anything: vector, hybrid and semantic search, skills, OCR (coming)
-
Configure semantic search, hybrid search, and vector search for grounding
Built in Episode 16: Index anything: vector, hybrid and semantic search, skills, OCR (coming)
-
Implement enrichment by using custom or built-in skills for text, images, and layout
Built in Episode 16: Index anything: vector, hybrid and semantic search, skills, OCR (coming)
-
Configure RAG ingestion flow, including documents and using optical character recognition (OCR)
Built in Episode 16: Index anything: vector, hybrid and semantic search, skills, OCR (coming)
-
Connect retrieval pipelines directly to workflows and agent tools
Built in Episode 9: Agent tools: APIs, knowledge, search, custom functions (coming)
Extract content from documents
-
Extract information by using multimodal pipelines that combine OCR, layout analysis, and field extraction
Built in Episode 17: Extract documents: OCR, layout, fields, analyzers (coming)
-
Produce clean, grounded representations to use with agents and RAG by using Content Understanding
Built in Episode 17: Extract documents: OCR, layout, fields, analyzers (coming)
-
Implement analyzers for generating structured or markdown outputs for downstream reasoning by using Content Understanding
Built in Episode 17: Extract documents: OCR, layout, fields, analyzers (coming)
How to use this guide
- Watch the episodes in order and build each skill yourself, from the code at each episode's tag in the course repository.
- After each skill, answer its practice questions here, or take a mixed practice test of 50, weighted as the exam is.
- Ask the AI-103 exam tutor about anything you got wrong: it answers with sources.
- Before you book, read the AI-103 exam FAQ (cost, registration, format, passing score) and take Microsoft's free practice assessment.
Independent course. Not affiliated with or endorsed by Microsoft.