NodnSix principles
- Fairness similar treatment
- Reliability and safety behaves as intended
- Privacy and security protect data
- Inclusiveness people can use it
- Transparency people can understand it
- Accountability people own it
Study guide · Flashcards & memory notes
Azure AI Fundamentals Flashcards & Memory Notes
40 recall cards for the facts AI-900 keeps asking you to separate: workloads and principles, learning tasks, vision and language services, and generative AI tools.
How to use these cards: Say the answer out loud before you flip the card, and mark it honestly. Revisit the cards you missed tomorrow rather than rereading them right away. Progress is kept only while this page is open. Exam AI-900 retired on June 30, 2026. For the full explanations, see the core notes.
Study mode
Choose a deck, flip each card, and mark what you already know.
You marked every card in this deck as known. Shuffle and run through it again, or choose another deck.
All flashcards
Select a question to reveal its answer.
Deck 1 · Workloads and responsible AI
Which job it is, and which principle is at stake.
1.1What does a computer vision workload do?
It interprets images or video, such as labeling a photo, finding an object, or reading text in a scan.
1.2What does a natural language processing workload do?
It interprets or produces human language, including text understanding and speech.
1.3What does a document processing workload extract?
Structured fields from forms, receipts, and invoices, such as a total, a date, and a vendor.
1.4What does a generative AI workload create?
New content from a prompt, such as a paragraph, code, or an image that did not already exist.
1.5What is fairness in a responsible AI solution?
People in similar situations are treated similarly. A model should not disadvantage a group that has the same qualifications.
1.6What is reliability and safety?
The system behaves as intended, including when it is unsure, and it avoids harmful output.
1.7What do privacy and security protect?
Personal data and access to the system. A prompt that leaks another customer's records fails this principle.
1.8What is inclusiveness?
People with different abilities can use the solution. A voice-only reply with no text alternative excludes someone who cannot hear it.
1.9What is transparency?
People can understand what the system does and where it is limited. Users should be able to tell that content was generated.
1.10What is accountability?
People are responsible for the system and its decisions. A model suggestion still needs an owner who can explain or override it.
Deck 2 · Machine learning
Tasks, columns, splits, and Azure Machine Learning.
2.1What does regression predict?
A number, such as a price or a temperature. Classification predicts a category.
2.2What does classification predict?
A category, such as spam or not spam. Regression predicts a number.
2.3What does clustering do?
It groups similar items when there is no label. You did not provide the group names in advance.
2.4What is a feature, and what is a label?
A feature is an input. A label is the value you want to predict. Clustering has no label.
2.5What is the difference between training data and validation data?
Training data is what the model learns from. Validation data is held out so you can measure generalization.
2.6What is a feature of deep learning?
Neural networks with many layers, useful when features such as pixels are hard to define by hand.
2.7What is a feature of the Transformer architecture?
Attention, which weighs how tokens relate, so the model is not limited to reading one word at a time.
2.8What does automated machine learning do?
It tries algorithms and settings for a task you define, then ranks models. You still provide the data.
2.9What is the difference between a compute instance and a compute cluster?
A compute instance is a development machine. A compute cluster is scalable compute for training.
2.10What is the difference between registering a model and deploying it?
Registering versions the model in the workspace. Deploying publishes it to an endpoint an application can call.
Deck 3 · Vision, language, and speech
The task, then the Azure service.
3.1What is the difference between image classification and object detection?
Classification labels the whole image. Object detection finds each object and returns a bounding box.
3.2What does optical character recognition return?
Text from an image, printed or handwritten. It does not judge sentiment.
3.3What is the difference between facial detection and facial analysis?
Detection finds a face and where it is. Analysis describes attributes, such as glasses or a smile. The skill list does not ask you to name the person.
3.4Which service analyzes images and can read text in them?
Azure AI Vision. Azure AI Face is for faces, not for receipt totals.
3.5What is the difference between key phrase extraction and entity recognition?
Key phrases are the main talking points. Entities are named things such as people, places, and dates.
3.6What does sentiment analysis return?
Whether text is positive, negative, neutral, or mixed. It does not translate the text.
3.7What is language modeling used for?
Learning how likely a sequence of words is, so a model can predict or generate text. It is not OCR.
3.8What is the difference between speech recognition and speech synthesis?
Recognition turns audio into text. Synthesis turns text into spoken audio.
3.9Which service handles speech to text, text to speech, and speech translation?
Azure AI Speech. Azure AI Language analyzes text that is already written.
3.10Which service handles sentiment, key phrases, and entities?
Azure AI Language. Do not use Azure AI Vision for the tone of a sentence.
Deck 4 · Generative AI on Azure
Models, risks, and the three Azure names.
4.1How is a generative model different from a classifier?
A generative model creates new content. A classifier only assigns a label to content that already exists.
4.2What is a prompt?
The instruction or input you give a generative model. The text it returns is the completion or response.
4.3What is a hallucination?
Fluent output that is not supported by facts. A content filter blocks harmful content. It does not make an answer true.
4.4What does grounding change?
You supply relevant source text with the prompt so the answer can stay closer to that source. It does not retrain the model.
4.5Name two common generative AI scenarios.
A chat assistant or copilot, and drafting text, code, or images. Labeling an existing photo is classification, not generation.
4.6What does Azure OpenAI provide?
OpenAI models in your Azure subscription, with Azure identity, networking, and content filters.
4.7What is Azure AI Foundry for?
Exploring models, trying prompts, and building a generative AI app. The May 2, 2025 guide uses this name. AI-901 later says Microsoft Foundry.
4.8What is in the Azure AI Foundry model catalog?
Foundation models from Microsoft, OpenAI, and other providers that you can review and deploy. It is not one vendor only.
4.9When would you use Azure Machine Learning instead of Azure OpenAI?
When you train or deploy your own model on your data, including automated machine learning. Azure OpenAI serves foundation-model completions.
4.10Which responsible AI issue is a confident but unsupported answer?
A hallucination. Telling users the text was generated is transparency. An owner who can override it is accountability.
Memory notes
Groupings that make the highest-yield AI-900 facts easier to recall.
Three learning tasks
- Regression a number
- Classification a category
- Clustering groups, no labels
Columns and splits
Features in, label out. Train to learn. Validate on rows the model has not seen.
Vision versus language
Vision labels, locates, or reads an image. Language extracts phrases, entities, or sentiment from text. Speech turns audio into text, or text into audio.
Services
- Azure AI Vision images and OCR
- Azure AI Face faces
- Azure AI Language text
- Azure AI Speech audio
Generative names
Azure OpenAI hosts OpenAI models. The model catalog lists more providers. Azure AI Foundry is where you try prompts and build.
Continue the Azure AI Fundamentals study path
Test what you have learned with a timed practice exam.
Overview
Retirement date, the 700 passing score, domain weights, and a study plan.
Core Notes
Reference notes for workloads, machine learning, vision, language, and generative AI.
Flashcards & Memory Notes
40 recall cards in four decks, plus memory notes for principles, models, and services.
Practice Exams
200 original questions with custom exams, explanations, and a score report by domain.
Related Tools
Useful companions while you study.
Password Generator
Generate strong random passwords with custom length, characters, symbols, and security options.
JSON YAML Converter
Convert structured data between JSON and YAML with formatting and validation.
Base64 Converter
Encode text and files to Base64 or decode Base64 data with UTF-8 support.
Diff Checker
Compare two text blocks and highlight added, removed, and changed content.
Legal Disclaimer
NodnWebTools provides general informational, educational, and convenience resources. Calculations, conversions, estimates, and learning materials may contain errors or become outdated. Financial, tax, medical, legal, and travel information is not professional advice. Verify important results and current requirements with qualified professionals or authoritative sources. Protect sensitive files and personal information, review each tool’s privacy limitations, and use only content you are authorized to process. You are responsible for how you use and share results. Study resources are independent and do not guarantee exam success or imply certification-provider endorsement. Microsoft, Azure, and related marks are trademarks of Microsoft Corporation. NodnWebTools is not affiliated with, endorsed by, or sponsored by Microsoft.