Skip to content

AI-900: Cheatsheet

<- Overview | Exam Guide ->


Domain Weights

DomainWeight
AI workloads and considerations15-20%
Machine learning on Azure15-20%
Computer vision workloads15-20%
NLP workloads15-20%
Generative AI workloads20-25%

Responsible AI

PrincipleRemember
FairnessAvoid bias; evaluate across groups
Reliability and safetyConsistent, safe behavior under expected conditions
Privacy and securityProtect data, access, prompts, and outputs
InclusivenessDesign for diverse users and accessibility needs
TransparencyExplain AI use, limits, and behavior
AccountabilityHuman ownership, review, audit, and escalation

ML Technique Lookup

ScenarioAnswer
Predict a numeric valueRegression
Predict a category or classClassification
Find natural groupings without labelsClustering
Learn complex image/audio/text patternsDeep learning
Language models and attentionTransformer architecture
Input column used for predictionFeature
Target value to predictLabel
Try multiple ML algorithms automaticallyAutomated ML

Vision Lookup

ScenarioAnswer
Assign label to entire imageImage classification
Locate objects in imageObject detection
Extract text from imageOCR
General image analysis/OCRAzure AI Vision
Face-specific detection/analysisAzure AI Face

NLP and Speech Lookup

ScenarioAnswer
Extract important termsKey phrase extraction
Extract people, places, dates, quantitiesEntity recognition
Determine positive/negative/neutral opinionSentiment analysis
Convert speech to textSpeech recognition
Convert text to speechSpeech synthesis
Convert between languagesTranslation
Text analytics serviceAzure AI Language
Speech input/output serviceAzure AI Speech

Generative AI Lookup

ScenarioAnswer
Generate text, summaries, code, imagesGenerative AI
Ground answers in private/fresh dataRAG
Numeric representation for similarityEmbedding
Model can process multiple input typesMultimodal model
OpenAI models through AzureAzure OpenAI Service
Build AI apps and manage model workflowsAzure AI Foundry
Discover and deploy modelsAzure AI Foundry model catalog

Quick Decision Rules

Input is image/video? -> Azure AI Vision, object detection, classification, OCR, or Face depending on output.

Input is text? -> Azure AI Language for analytics, Azure OpenAI for generation, Translator for language conversion.

Input/output is audio? -> Azure AI Speech.

Need custom ML lifecycle? -> Azure Machine Learning.

Need generative AI app platform? -> Azure AI Foundry.

Need OpenAI models with Azure controls? -> Azure OpenAI Service.


Retirement Reminder

AI-900 is retiring on June 30, 2026. If studying after that date, check AI-901: Azure AI Fundamentals instead.


<- Overview | Exam Guide ->