AI-900: Microsoft Azure AI Fundamentals Course - May 2022

Learn to create a free Azure subscription with 200 USD credit for 30 days and 25 always-free services, plus 12-month storage and database allowances.

Learn to keep the Azure portal free after the 12-month trial using Microsoft Learn sandboxes; switch directories and run up to four hours per sandbox, up to 10 daily.

Explore the basics of artificial intelligence, including prediction, forecasting, anomaly detection, vision, natural language processing, and conversational AI. Learn guiding principles—fairness, reliability and safety, privacy and security, accountability, and transparency.

Explore how artificial intelligence differs from natural intelligence and how an agent, environment, action, and goal define ai. See practical industry applications from customer service to healthcare and finance.

Differentiate prediction and forecasting using data from previous events to forecast future outcomes. Explore stock prices, monthly bill, flight arrivals, weather, and electricity consumption as examples.

Learn how anomaly detection uses a machine learning model to identify values outside the expected range over time, enabling real-time or historical analysis and possible preemptive action.

Explore computer vision workloads by showing how images and videos yield tags, descriptions, and confidence scores. Learn to read text and handwriting with OCR and recognize landmarks or celebrities.

Explore conversational AI workloads that let an agent converse with humans using natural language processing; discover chatbots for customer support, reservations, health care consultations, and home automation as digital assistants.

Explore the six guiding principles of responsible artificial intelligence—fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability—and how they shape ethical model design.

Learn the fairness principle of responsibility in AI, ensuring equal, unbiased treatment across loans, medical treatment, and employment by using diverse data and reducing model bias.

Developers must protect sensitive data by ensuring strong privacy and security, complying with standards and laws, and transparently communicating data collection, use, and storage with consumers.

The lecture highlights the inclusiveness guiding principle, stating that intelligent technology should benefit everyone regardless of ability or identity. It emphasizes designing voice, mobility, and vision solutions to remove barriers.

Explain how your application makes decisions, outline the rationale and limitations, and enable developers to override decisions when necessary to ensure transparency.

Accountability means you are responsible for an application's decisions and can override them. Design within governments and organizational principles to meet ethical standards, safeguard civil liberties, and review facial recognition.

Explore the three main machine learning types—classification, regression, and clustering—along with supervised versus unsupervised learning, and a brief note on reinforcement learning.

Learn to distinguish features from the label, apply feature selection to remove irrelevant data and redundancy, and use feature engineering to create informative new features that boost model performance.

Explore machine learning algorithms across classification, clustering, and regression, with examples like logistic regression, boosted trees, and k-means, and learn how training time, accuracy, and features influence choice.

Create and access a machine learning workspace in the Azure portal, then design, train, and deploy models with the drag-and-drop studio, or with Python, R, or VS Code.

Design and deploy a regression model in Azure Machine Learning Studio by cleaning and normalizing data, splitting 70/30, training with linear regression, and evaluating and deploying a real-time endpoint.

Stop idle compute instances to avoid costs, set cluster min zero and max one, and delete new datasets to minimize storage charges; keep notebooks and pipelines.

Train a two-class logistic regression classification model on a tabular diabetes dataset, normalize features, split data 70/30, evaluate with accuracy, precision, recall, and F1 score, and deploy real-time inference pipeline.

Learn automated machine learning for regression, compare algorithms with metrics like normalized RMSE, and deploy the best model for bike rental data via a REST endpoint and Python notebook.

Delete the compute resources and prevent unintended costs after these machine learning studio demos, and leave notebooks, designer, and dataset experiment intact.

Master optical character recognition (ocr) to extract printed and handwritten text from input images and documents, enabling digitized text from notes, invoices, medical records, historical and official documents, and checks.

Demonstrates setting up a computer vision service, retrieving endpoint and keys, and running image analysis in a notebook to describe images and generate tags.

Learn how to build a domain-specific model with the Custom Vision Service, using classification and object detection, plus separate training and prediction resources.

Explore Azure face service to locate and analyze faces in images or videos, returning attributes like age, emotion, glasses, and mask, with detection and recognition operations like verify and identify.

Encourage learners to rate and review the course using the top-right feedback controls, save progress with save and continue, and edit ratings or reviews as needed.

Learn the objectives of the fourth module on natural language processing, covering key features such as key phrase extraction, entity recognition, sentiment analysis, language modeling, speech recognition, and translation.

Explore natural language processing by showing how computers interpret human language, infer intent, and act on commands, with examples like turning on all the lights and entity recognition.

Explore language modeling and how it enables language understanding by identifying intent and entities to take action, with applications in home automation, customer support, reservation systems, and more.

Explore the translation API in Azure Cognitive Services, enabling text and speech translation across languages, including real-time speech-to-speech translation, document or webpage translation, and email translation.

Explore how to implement NLP features with Azure tools, including text analytics, language understanding, speech, and translation services, and learn when to use cognitive service versus individual services.

Learn how the text analytics service performs language detection, key phrase extraction, named entity recognition, and sentiment analysis, and how to call it from canvas apps and notebooks.

Explore the Microsoft Azure speech service for real-time and batch speech to text, text to speech, and speech translation. Use synthesis markup language to adjust pitch, pause, and pronunciation.

Learn to build a language understanding model that identifies intents and entities from utterances to control devices such as lights and fans, train, test, and publish to a prediction endpoint.

Explore conversational AI use cases across web chat bots, personal digital assistants, and telephone voice systems, including real-time knowledge base responses, calendar management, and automated customer support.

Learn to build a conversational layer over a knowledge base with QnA Maker and Bot Framework, training and publishing static information for accurate answers.

Demonstrate building a q&a maker with a knowledge base, training, testing, and publishing. Deploy a bot via web chat and the Bot Framework, demonstrating web app and channel integration.

Master AI-900 exam strategies for a 60-minute, 45–55-question test. Identify question types, use elimination and keywords like SMB protocol to maximize partial-credit scoring.