[2025] Microsoft Azure AI Fundamentals | AI-900

Explore Microsoft's responsible AI framework, detailing six principles: fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability, and how they guide ethical AI design, deployment, and governance.

Machine learning teaches computers to learn from data through training, creating a model that performs inferencing on new data. Features and labels drive predictions, and data quality and quantity matter.

Explore the two major types of machine learning: supervised learning with labeled data, including regression and classification, and unsupervised learning with unlabeled data, such as clustering.

Explore regression in machine learning by predicting numeric labels from features using a linear regression model trained on labeled data and evaluated on unseen validation data, with iterative parameter tuning.

Explore regression metrics including mean absolute error, mean squared error, root mean squared error, and the coefficient of determination to assess prediction accuracy on a validation set for house prices.

Explore regression training by selecting and preparing features, choosing the right algorithm, and tuning hyperparameters through iterative retraining to optimize evaluation metrics.

Explore binary and multi-class classification, train models with features and labels, and evaluate using confusion matrices and metrics.

Evaluate multiclass classification by computing per-class binary metrics from the confusion matrix—true positives, false positives, false negatives, true negatives—and aggregate into accuracy, recall, precision, and F1, illustrated with penguin species.

Explore binary classification evaluation metrics, including accuracy, recall, precision, and F1, illustrated with a diabetes example and the confusion matrix.

Perform clustering, an unsupervised machine learning method that groups observations by similarities using features and centroids. Iterate by assigning points to the nearest centroid and updating centers until convergence.

Train and evaluate a bike rental regression model using seasonal and meteorological features in Azure Machine Learning with AutoML, then deploy a real-time endpoint.

Explore Azure AI services as ready-to-use AI building blocks—image recognition, natural language processing, speech, and AI powered search—accessible via API with single or multi-service resources, SDKs, and studios.

Explore how computer vision enables machines to interpret visual information by treating images as pixel grids, with color channels, filters, and convolution techniques like the Laplace edge detector.

Train convolutional neural networks on labeled images to extract features with convolutional filters, create feature maps, flatten them, and classify objects with softmax probabilities.

Explore how transformers and multimodal models transform computer vision by encoding text and images into embeddings, enabling tasks like image classification, object detection, and image captioning with Azure AI Vision.

Create and configure Azure AI services, explore computer vision capabilities like video retrieval and summary, object detection, and ocr, and learn to train custom vision models for your domain.

Explore face detection, analysis, and recognition, using bounding boxes and facial features for training images. See how Azure AI Vision, AI Video Indexer, and AI Face provide these capabilities.

Navigate the Azure portal to create a face Api resource, explore face detection in images, view JSON outputs, and learn about facial attributes and masks before cleaning up the resource.

Discover how optical character recognition blends computer vision and natural language processing to extract machine readable text from images, PDFs, and TIFFs using the Azure read API.

Create an Azure AI OCR resource and use Vision Studio to extract text from images, view words and lines with bounding boxes in JSON, and explore REST API or SDK.

Explore natural language processing fundamentals, including tokenization, corpus analysis, and frequency analysis. Learn how semantic language models and embeddings power sentiment, summarization, and text classification with logistic regression.

Explore Azure AI language features for natural language processing on unstructured text, including named entity recognition with entity linking, pii detection, language detection, sentiment analysis, summarization, and key phrase extraction.

Create and connect a language resource in Azure AI services, then use Language Studio to extract named entities, classify texts, and analyze sentiment with JSON results and code examples.

Discover how Azure language's custom question answering builds a knowledge base from your data to answer questions. See how Azure Bot Service and channels enable 24/7 chatbots and help centers.

Create a travel knowledge base with azure ai studio’s custom question answering, train and test it from an FAQ, deploy a bot, and connect channels for live q&a.

Perform cleanup by deleting the language resource, theAzure bot, and the test resource group, including app service, manage identity, and app service plan, after confirming deletion.

Explore conversational language understanding by defining utterances, entities, and intents with examples like turning on the fan or light, and explore Azure AI language for authoring, training, and predicting.

Build a smart home assistant by creating a language service, defining intents like switch on/off, labeling device entities, and training, deploying, and testing the model in Azure.

Azure AI Speech offers speech to text and text to speech capabilities with pre-built and custom models to transcribe audio, identify speakers, and create custom voices.

learn how to deploy and connect an azure ai speech resource, explore real-time speech to text, batch transcription, text-to-speech synthesis, and call center transcription with insights and PII redaction.

Explore Azure AI translator and Azure AI speech, powered by neural machine translation to translate text and speech across more than 130 languages, with multi-language output and custom translations.

Explore building a translation service in Azure AI with translator creation, testing German to English translations, and using text and document translation endpoints to power multilingual apps.

Master Azure ai document intelligence to digitize text with OCR, extract data from invoices and forms, and deploy pre-built or custom models for end-to-end workflows.

Create a document intelligence resource in the Azure portal, configure the subscription and resource group, and explore document analysis with bounding boxes using prebuilt and custom models.

Explore knowledge mining across structured, semi-structured, and unstructured data to uncover insights. Learn how azure ai search indexes, enriches, and retrieves content from diverse sources.

Build a knowledge mining solution with Azure AI Search by turning Forth Coffee customer reviews into enriched insights via location, sentiment, and key phrases.

Explore how generative AI creates original content from prompts using transformer language models, encoder and decoder architectures, and tokenization, and learn to use Azure foundation models with fine-tuning.

Explore copilots as contextual AI assistants embedded in apps that boost productivity by automating tasks and generating drafts, with options to use off-the-shelf, extend, or build custom copilots.

Explore Azure AI Foundry, a unified workshop uniting machine learning tools, Azure OpenAI integration, and specialized AI services, organized into AI hubs and projects for collaborative development and deployment.

Adopt a four-stage, accountable approach to responsible generative AI. Identify, measure, mitigate, and operate risks aligned with the NIST AI risk management framework.

Learn how to manage Azure AI Studio costs with budgets and alerts, then build and deploy AI solutions using hubs, model catalog, prompt catalog, AI services, and prompt flow.

Learn to create and apply content filters in Azure AI Foundry, configure input/output safety categories and severity, and use a block list with deployed models.