AI-103 cheat sheet

Exam at a glance

Domain weightings

How much of the exam each domain covers. Spend your study time in proportion — the heavier the domain, the more questions you'll see.

Key terms

Microsoft Foundry

Microsoft Foundry is Microsoft's unified platform for building, deploying, and managing generative AI apps and agents on Azure. AI-103 centers on developing solutions with Foundry projects, SDKs, and connectors.

Foundry Agent Service

The Foundry Agent Service is the Microsoft Foundry capability for creating and running AI agents with tools, knowledge, and memory. AI-103 requires building agents and orchestrating multi-agent solutions with it.

Agent

An agent is an AI system that pursues goals by reasoning over tools, knowledge, and conversation memory rather than answering a single prompt. AI-103 covers defining agent roles, tool schemas, and approval-gated autonomous workflows.

RAG

RAG (retrieval-augmented generation) is a pattern that grounds a language model's responses in retrieved data to improve accuracy and reduce fabrication. AI-103 requires implementing RAG in applications and tuning its retrieval.

Grounding

Grounding is the practice of anchoring a generative model's output in trusted source data so responses are factual and cite evidence. AI-103 covers grounding models in your data through retrieval pipelines.

Vector search

Vector search retrieves content by comparing embedding vectors for semantic similarity rather than exact keywords. AI-103 covers configuring semantic, hybrid, and vector search for grounding.

Azure AI Search

Azure AI Search is the Azure service that indexes content and serves semantic, hybrid, and vector queries for grounding AI applications. AI-103 uses it to build retrieval and grounding pipelines.

Function calling

Function calling lets a language model invoke defined tools or APIs by returning structured arguments the application executes. AI-103 covers integrating function-calling into agents alongside retrieval and memory.

Model catalog

The model catalog is the Microsoft Foundry gallery where you browse, compare, and deploy foundation models — including LLMs, small language models, and multimodal models — from Microsoft, OpenAI, and partners. AI-103 covers choosing the appropriate model for a task from the catalog.

Azure OpenAI

Azure OpenAI in Foundry Models provides access to OpenAI models such as the GPT chat models and the gpt-image image-generation models through Azure with enterprise security and networking. AI-103 covers provisioning, deploying, and consuming these models.

Content Understanding

Azure Content Understanding in Foundry Tools extracts structured information from documents, images, video, and audio for grounding and analysis. AI-103 uses it for multimodal understanding and information extraction.

Document Intelligence

Azure Document Intelligence is the Azure service that extracts text, key-value pairs, and tables from documents using prebuilt or custom models. AI-103 covers building document-extraction pipelines with it.

OCR

OCR (optical character recognition) converts text in images and scanned documents into machine-readable text. AI-103 uses OCR within RAG ingestion and multimodal extraction pipelines.

Prompt shields

Prompt shields are an Azure content-safety capability that detects and blocks prompt-injection and jailbreak attempts against a model. AI-103 covers preventing harmful behavior with prompt shields and harm detection.

Responsible AI

Responsible AI is the set of principles and controls — fairness, safety, privacy, transparency, and accountability — applied across an AI solution. AI-103 requires implementing content filters, guardrails, and governance to uphold it.