Microsoft Azure AI Fundamentals

Students will explore different AI workload types and match Azure services to real-world scenarios.

1. Review the descriptions and use cases of computer vision, NLP, document processing, and generative AI workloads.
2. Access the Azure AI services documentation and list the primary service(s) associated with each workload type.
3. Using the Azure portal, identify sample scenarios and classify which Azure AI workload they correspond to.
4. Prepare a short summary of why each Azure service fits the identified workload.

Expected Outcome: Students will be able to distinguish AI workloads and associate them with appropriate Azure services.

Analyze provided real-world scenarios and identify suitable Azure AI workload types and services.

1. Read provided scenario descriptions (e.g., automated document processing, chatbot customer support).
2. For each scenario, select the most appropriate AI workload type (computer vision, NLP, document processing, generative AI).
3. Identify which Azure AI service(s) would support the scenario and explain your choice.
4. Submit your analysis for peer review or instructor feedback.

Expected Outcome: Students will practice applying AI workload knowledge to practical Azure scenarios.

Hands-on use of the Custom Vision service to create an image classification project.

1. Sign in to the Azure portal and navigate to the Custom Vision service.
2. Create a new project with the classification type set to multiclass.
3. Upload sample images into at least three categories (e.g., animals, vehicles, fruits).
4. Train the model and evaluate its performance using the built-in evaluation metrics.
5. Test the model with new images and observe predictions.

Expected Outcome: Students will understand how to set up and train image classification models using Azure Custom Vision.

Create an object detection model using Azure Custom Vision to detect objects in images.

1. Log in to the Azure Custom Vision portal and start a new object detection project.
2. Upload images containing multiple objects and use bounding boxes to tag objects.
3. Train the object detection model and review the precision and recall metrics.
4. Test the model with new images and interpret the detection results.

Expected Outcome: Students will learn how to build and evaluate object detection models with Azure Custom Vision.

Use Azure AI Language Studio to analyze sentiment of sample text data.

1. Access Azure AI Language Studio and select the sentiment analysis feature.
2. Input different sample sentences expressing positive, negative, and neutral sentiments.
3. Run the sentiment analysis and record the sentiment scores and labels.
4. Examine the confidence scores and explain how they affect interpretation.

Expected Outcome: Students will gain practical experience running sentiment analysis and interpreting results using Azure AI Language.

Experiment with Azure Speech service to convert speech to text and text to speech.

1. Navigate to the Azure Speech Studio and access Speech to Text and Text to Speech samples.
2. Record or upload a short audio clip and transcribe it using Speech to Text.
3. Use Text to Speech to generate spoken audio from a typed sentence.
4. Compare the results and note any challenges or limitations.

Expected Outcome: Students will understand how to use Azure Speech services for recognition and synthesis tasks.

Use Azure Document Intelligence to perform OCR on scanned documents and extract structured information.

1. Create or use an existing Azure Document Intelligence resource in the Azure portal.
2. Upload sample scanned documents or images containing forms or printed text.
3. Run the OCR feature and review the extracted text and layout information.
4. Export and analyze the extracted data for accuracy.

Expected Outcome: Students will learn to extract text and form data from documents using Azure Document Intelligence OCR.

Train a custom model with Azure Document Intelligence to extract specific fields from forms.

1. Prepare a dataset of sample forms with labeled fields (e.g., invoices, surveys).
2. Upload the labeled forms to Azure Document Intelligence's custom model training interface.
3. Train the custom model and test it on new form data.
4. Evaluate extraction accuracy and adjust training data as needed.

Expected Outcome: Students will understand how to create and deploy custom document processing models with Azure Document Intelligence.

Use Azure AI Foundry portal to interact with GPT models for text generation.

1. Access Azure AI Foundry portal and sign in with your Azure account.
2. Select a GPT-based model and experiment with prompt inputs to generate text.
3. Adjust parameters like temperature and max tokens to observe changes in output.
4. Document the use cases best suited for generative AI models.

Expected Outcome: Students will gain hands-on experience with generative text models on Azure OpenAI service.

Leverage Azure OpenAI DALL-E model to generate images from text prompts.

1. Navigate to Azure AI Foundry portal and select the DALL-E model.
2. Input descriptive text prompts to generate corresponding images.
3. Experiment with prompt variations to refine image outputs.
4. Save and review generated images for quality and relevance.

Expected Outcome: Students will understand how to generate images using generative AI models on Azure.

Use Azure Machine Learning fairness dashboard to evaluate model bias.

1. Train or use an existing classification model in Azure Machine Learning Studio.
2. Access the fairness dashboard to analyze model predictions across sensitive attributes.
3. Identify any detected bias or fairness issues in the model.
4. Experiment with mitigation techniques like reweighting or sampling and observe effects.

Expected Outcome: Students will learn to detect and mitigate bias in Azure ML models using built-in tools.

Use Azure ML SDK to calculate fairness metrics programmatically.

1. Set up an Azure ML workspace and prepare a trained classification model.
2. Write Python code using Azure ML SDK to compute fairness metrics like disparate impact and equal opportunity difference.
3. Interpret the metric results to assess fairness.
4. Modify dataset or model parameters to improve fairness metrics.

Expected Outcome: Students will gain coding experience with Azure ML SDK to evaluate fairness quantitatively.

Set up monitoring for an AI model deployment to track reliability and detect anomalies.

1. Deploy a machine learning model as a web service in Azure ML.
2. Configure Azure Application Insights to collect telemetry data from the endpoint.
3. Simulate requests and observe metrics like latency, failure rate, and exceptions.
4. Analyze logs to identify reliability or safety issues.

Expected Outcome: Students will understand how to monitor AI models for reliability and safety using Azure tools.

Conduct adversarial testing on an Azure ML model to evaluate robustness.

1. Obtain or train a classification model in Azure ML workspace.
2. Use adversarial example generation techniques to create perturbed inputs.
3. Test model predictions on these adversarial inputs.
4. Document model performance degradation and propose mitigation strategies.

Expected Outcome: Students will learn about safety challenges and test model robustness in Azure ML.

Use Azure Data Factory and Azure Purview to anonymize sensitive data before model training.

1. Create a pipeline in Azure Data Factory to ingest raw data containing PII.
2. Apply data masking or pseudonymization techniques using Data Factory transformations.
3. Catalog and classify data with Azure Purview to track privacy compliance.
4. Verify anonymized dataset suitability for training AI models.

Expected Outcome: Students will understand methods to protect privacy in AI datasets using Azure services.

Set up RBAC policies to secure access to Azure AI services and data.

1. Identify Azure resources used for an AI workload (e.g., Azure ML workspace, Azure AI services).
2. Define roles and assign least privilege access using Azure RBAC in the portal.
3. Test access permissions by attempting operations with different user roles.
4. Document security best practices for AI workloads.

Expected Outcome: Students will learn to enforce security controls on AI resources via Azure RBAC.

Apply Azure ML interpretability features to explain classification model decisions.

1. Train or select a pre-trained classification model in Azure ML workspace.
2. Use the Azure ML interpretability dashboard or SDK to compute feature importance and SHAP values.
3. Analyze individual prediction explanations and overall model transparency.
4. Share findings emphasizing how transparency supports inclusiveness.

Expected Outcome: Students will experience how to improve AI transparency using Azure ML interpretability tools.

Prepare a dataset ensuring diverse representation to improve inclusiveness in AI training.

1. Analyze an existing dataset for demographic or categorical imbalances.
2. Augment the dataset by adding or synthesizing examples from underrepresented groups.
3. Use Azure ML Data Labeling or Data Preparation tools to balance the dataset.
4. Train a model on the balanced dataset and compare fairness metrics.

Expected Outcome: Students will learn methods to improve AI inclusiveness by dataset balancing using Azure tools.

Use Azure Policy to enforce compliance rules on AI resources and deployments.

1. Review organizational AI governance requirements (e.g., data usage, model approval).
2. Create and assign Azure Policy definitions to restrict AI resource configurations.
3. Test policy enforcement by attempting to deploy non-compliant resources.
4. Document audit logs and compliance reports generated by Azure Policy.

Expected Outcome: Students will understand how to implement accountability via Azure governance tools.

Set up auditing to track AI model changes and usage using Azure Monitor.

1. Enable diagnostic logging for Azure ML workspace and AI services.
2. Configure Azure Monitor and Log Analytics workspace to collect logs.
3. Query logs to identify model deployments, modifications, and access events.
4. Generate reports demonstrating accountability over AI lifecycle.

Expected Outcome: Students will gain skills in auditing AI systems for accountability with Azure monitoring tools.

Create and train a regression model to predict housing prices using Azure ML designer.

1. Import a housing dataset with features like size, location, and price.
2. Select regression algorithms such as Linear Regression or Decision Forest Regression.
3. Train and evaluate the model using metrics like RMSE and R-squared.
4. Test the model predictions on new data samples.

Expected Outcome: Students will build and evaluate regression models on Azure ML.

Use Azure Automated ML to create a forecasting model for sales data.

1. Upload a time series dataset with date and sales columns.
2. Configure Automated ML for regression with time series forecasting settings.
3. Run the experiment and review the best model selected.
4. Deploy the model and generate forecasts for future periods.

Expected Outcome: Students will understand regression forecasting scenarios using Azure Automated ML.

Build a binary classification model to predict customer churn.

1. Import a customer dataset with features and churn labels.
2. Select and configure classification algorithms like Logistic Regression or Decision Trees.
3. Train and evaluate the model using accuracy and AUC metrics.
4. Test the model on unseen data for prediction.

Expected Outcome: Students will build binary classification models on Azure ML platform.

Use Azure Automated ML to classify images into multiple categories.

1. Prepare a labeled dataset with multiple image categories.
2. Configure Automated ML for multi-class classification.
3. Train the model and review classification metrics like precision, recall, and F1 score.
4. Deploy the model and perform inference on new images.

Expected Outcome: Students will learn to build multi-class classification models with Azure Automated ML.

Apply clustering algorithms to segment customers based on purchasing behavior.

1. Load a customer dataset with purchase frequency and amount features.
2. Use the K-Means clustering module in Azure ML designer.
3. Train the model and visualize clusters.
4. Interpret clusters to identify distinct customer segments.

Expected Outcome: Students will understand clustering and its application in customer segmentation using Azure ML.

Use anomaly detection to identify unusual data points in a dataset.

1. Import a dataset containing normal and anomalous records.
2. Configure Azure ML anomaly detection algorithms like Isolation Forest.
3. Train and evaluate the model's ability to detect anomalies.
4. Analyze detected anomalies and their impact.

Expected Outcome: Students will learn to implement clustering-based anomaly detection on Azure ML.

Practice selecting features and labels from datasets and splitting into training and validation sets.

1. Choose a sample dataset relevant to a prediction task.
2. Identify and separate feature columns and label column.
3. Use Azure ML Data Prep SDK or Designer to split data into training and validation sets.
4. Document the rationale behind feature and label selection.

Expected Outcome: Students will understand dataset preparation concepts essential for ML training on Azure.

Create new features and preprocess data to improve model input quality.

1. Load a dataset into Azure ML designer or notebook environment.
2. Apply transformations like normalization, encoding categorical variables, and creating interaction features.
3. Analyze the impact of engineered features on model training.
4. Prepare the final dataset for model consumption.

Expected Outcome: Students will gain practical skills in feature engineering and dataset refinement in Azure ML.

Implement a simple deep learning model using TensorFlow or PyTorch on Azure ML compute.

1. Set up an Azure ML compute instance and notebook environment.
2. Load a dataset (e.g., MNIST) and preprocess it.
3. Define and train a neural network model using Keras or PyTorch.
4. Evaluate model accuracy and visualize training metrics.

Expected Outcome: Students will understand deep learning fundamentals and Azure ML support for neural networks.

Review and compare CNN, RNN, and autoencoder architectures via Azure AI demos.

1. Access Azure ML sample notebooks demonstrating CNN, RNN, and autoencoder models.
2. Run training scripts and observe differences in architecture and application.
3. Summarize use cases best suited for each architecture.
4. Discuss how Azure supports these models with compute and tools.

Expected Outcome: Students will recognize common deep learning architectures and their Azure implementations.

Interact with transformer-based language models to understand attention mechanisms.

1. Access Azure AI Foundry portal and select a transformer-based GPT model.
2. Input prompts illustrating attention to context (e.g., question answering).
3. Analyze how the model handles long context dependencies.
4. Document observations about transformer capabilities.

Expected Outcome: Students will gain insight into transformer architecture features using Azure OpenAI models.

Use visualization tools to explore attention weights in transformer models.

1. Load pre-trained transformer model outputs (e.g., from Huggingface on Azure ML).
2. Use visualization libraries (like BertViz) to display attention maps for sample inputs.
3. Interpret how attention focuses on different input tokens.
4. Relate observations to transformer architecture principles.

Expected Outcome: Students will understand the attention mechanism's role in transformers via Azure ML tooling.

Analyze various computer vision use cases and map them to Azure tools.

1. Review workload types: classification, detection, OCR, video processing.
2. Research Azure services supporting each workload (e.g., Custom Vision, Document Intelligence, Video Indexer).
3. Create a mapping table of workload types to Azure services.
4. Present findings with example scenarios.

Expected Outcome: Students will be able to classify computer vision workloads and identify Azure tools accordingly.

Use Azure Computer Vision REST API to perform various vision tasks on sample images.

1. Set up Azure Computer Vision resource and obtain API keys.
2. Use Postman or Azure SDK to call image analysis, OCR, and object detection endpoints on sample images.
3. Capture and interpret JSON responses for each API call.
4. Summarize the suitability of each API for different workloads.

Expected Outcome: Students will gain practical skills using Azure Computer Vision APIs for diverse workloads.

Create and deploy an image classification project using Azure Custom Vision Studio.

1. Create a new Custom Vision project with classification domain.
2. Upload labeled images and tag them appropriately.
3. Train the classification model and evaluate performance.
4. Publish the model and test predictions via REST API.

Expected Outcome: Students will learn image classification model lifecycle on Azure Custom Vision.

Enhance an existing image classification model via incremental training and dataset refinement.

1. Analyze initial model performance and identify weak categories.
2. Add more training images or correct mislabeled samples.
3. Retrain the model and compare new performance metrics.
4. Document improvements and lessons learned.

Expected Outcome: Students will practice refining image classification models using Azure Custom Vision.

Create an object detection project and label bounding boxes on images.

1. Initiate a Custom Vision object detection project.
2. Upload images containing multiple object types.
3. Label objects with bounding boxes and train the model.
4. Evaluate detection precision and recall, then test with new images.

Expected Outcome: Students will understand object detection workflows on Azure Custom Vision.

Study common challenges in object detection and explore how Azure features address them.

1. Research challenges like occlusion, small object detection, and real-time processing.
2. Review Azure Custom Vision features such as image augmentation and domain-specific models.
3. Write a brief report summarizing challenges and Azure’s mitigating capabilities.
4. Suggest improvements or best practices for object detection projects.

Expected Outcome: Students will gain insights into object detection challenges and Azure solutions.

Perform OCR on scanned documents using Azure Computer Vision OCR API.

1. Set up Azure Computer Vision resource and retrieve API keys.
2. Use Azure SDK or REST API to submit images for OCR processing.
3. Extract and review text output, including layout and language identification.
4. Compare OCR results on different image qualities.

Expected Outcome: Students will learn to extract text from images using Azure OCR capabilities.

Use Document Intelligence to extract key-value pairs and tables from forms.

1. Create a Document Intelligence resource in Azure.
2. Upload sample invoices or forms to the Document Intelligence Studio.
3. Use prebuilt or custom models to extract structured data.
4. Validate extraction accuracy and export results.

Expected Outcome: Students will gain experience with structured document processing using Azure Document Intelligence.

Upload videos to Azure Video Indexer and analyze extracted metadata.

1. Create an Azure Video Indexer account and upload a sample video.
2. Wait for indexing to complete and explore generated insights like faces, speech-to-text, and sentiment.
3. Use the Video Indexer API to retrieve metadata programmatically.
4. Summarize how these insights can be used in applications.

Expected Outcome: Students will learn how to use Azure Video Indexer for rich video content analysis.

Configure custom video indexing parameters and evaluate impact on results.

1. Upload videos with different content types (e.g., interviews, events).
2. Modify indexing settings such as language, transcription accuracy, and face detection sensitivity.
3. Compare outputs and note improvements or trade-offs.
4. Document best practices for video indexing configuration.

Expected Outcome: Students will understand customization options for Azure Video Indexer.

Explore common NLP tasks and identify Azure services suited for each.

1. List NLP tasks such as sentiment analysis, entity recognition, translation, and speech-to-text.
2. Research Azure AI Language and Speech services to find matching features.
3. Create a table linking NLP workload types to Azure services and sample use cases.
4. Present findings to peers or instructor.

Expected Outcome: Students will be able to classify NLP workloads and associate them with Azure AI services.

Use Azure AI Language Studio to run multiple NLP tasks on provided texts.

1. Access Azure AI Language Studio and input sample texts.
2. Run tasks such as key phrase extraction, language detection, and entity recognition.
3. Review and interpret the results for each task.
4. Summarize differences and application scenarios.

Expected Outcome: Students will gain hands-on experience with multiple NLP capabilities on Azure.

Use Azure AI Language sentiment analysis to classify social media text sentiment.

1. Gather sample social media comments or tweets.
2. Submit text data to Azure AI Language sentiment analysis endpoint.
3. Interpret sentiment scores and labels for each comment.
4. Visualize distribution of sentiments using charts.

Expected Outcome: Students will learn to apply sentiment analysis on real-world text data using Azure.

Train a custom sentiment analysis model in Azure Language Studio for domain-specific texts.

1. Collect a labeled dataset of domain-specific text with sentiment labels.
2. Use Azure Language Studio to create a custom sentiment analysis project.
3. Train and evaluate the model on test data.
4. Compare performance to prebuilt sentiment analysis and document improvements.

Expected Outcome: Students will understand how to customize sentiment analysis models in Azure.

Use Azure AI Language key phrase extraction on customer feedback texts.

1. Collect sample customer feedback or reviews.
2. Submit texts to Azure AI Language key phrase extraction API.
3. Review extracted key phrases and their relevance.
4. Group feedback topics based on key phrases.

Expected Outcome: Students will learn to extract meaningful key phrases to summarize text data.

Analyze texts by extracting key phrases and associated sentiment to find opinion trends.

1. Submit a batch of texts to key phrase extraction and sentiment analysis endpoints.
2. Map sentiments to extracted key phrases.
3. Identify key topics with positive or negative sentiments.
4. Present findings as actionable insights.

Expected Outcome: Students will understand how to combine NLP tasks for richer text analytics.

Use Azure AI Language service to detect languages in a mixed-language text dataset.

1. Prepare or obtain a dataset containing sentences in various languages.
2. Submit texts to the Azure language detection API.
3. Review detected languages and confidence scores.
4. Identify challenges with short or ambiguous texts.

Expected Outcome: Students will gain practical experience in detecting languages on Azure AI Language service.

Create a text processing workflow that routes texts based on detected language.

1. Use Azure Logic Apps or Azure Functions to call language detection API.
2. Based on detection, route text to appropriate translation or NLP services.
3. Test pipeline with multi-language inputs.
4. Document architecture and benefits for multi-language applications.

Expected Outcome: Students will learn to build multi-language processing solutions using Azure AI Language.

Use Azure AI Language NER to identify entities in news text data.

1. Collect sample news article texts.
2. Submit texts to Azure NER API via Language Studio or SDK.
3. Review identified entities such as people, locations, and organizations.
4. Analyze entity frequency and relevance.

Expected Outcome: Students will understand named entity recognition and its application on Azure.

Design a simple app that extracts entities from customer emails to automate ticket routing.

1. Use Azure Logic Apps or Functions to process incoming emails.
2. Call Azure NER API to extract relevant entities (e.g., product names, issue types).
3. Map entities to support teams and route tickets accordingly.
4. Test with sample emails and refine entity mappings.

Expected Outcome: Students will apply NER to automate business processes using Azure AI services.

Use Azure Translator API to translate text between multiple languages.

1. Set up Azure Translator resource and obtain API keys.
2. Write code or use Azure Translator Studio to translate text samples.
3. Experiment with different source and target languages.
4. Evaluate translation quality and latency.

Expected Outcome: Students will learn to use Azure Translator for machine translation tasks.

Build a simple chatbot that translates user input into English before processing.

1. Use Azure Bot Service to create a chatbot.
2. Integrate Azure Translator API to detect and translate user messages.
3. Process translated text and generate responses.
4. Test chatbot with inputs in various languages.

Expected Outcome: Students will understand how to incorporate translation services into AI applications.

Interact with GPT models to generate text based on prompts.

1. Access Azure AI Foundry portal and select a GPT model.
2. Input different prompts related to storytelling, summarization, or question answering.
3. Adjust generation parameters and observe output changes.
4. Discuss model capabilities and limitations.

Expected Outcome: Students will gain hands-on experience with generative text models on Azure.

Use Azure OpenAI models to create images from text prompts and generate code snippets.

1. In Azure AI Foundry portal, experiment with DALL-E to generate images from descriptive prompts.
2. Use GPT-4o to generate code based on natural language instructions.
3. Evaluate generated outputs for accuracy and creativity.
4. Document appropriate use cases for each model.

Expected Outcome: Students will learn practical use of generative AI models for content and code generation.

Create a conversational AI chatbot leveraging GPT models for natural dialogue.

1. Set up Azure OpenAI resource and obtain API keys.
2. Develop a chatbot application that sends user input as prompts to GPT model.
3. Implement conversation context handling and response generation.
4. Test chatbot with typical user queries.

Expected Outcome: Students will build a generative AI chatbot using Azure OpenAI services.

Research and present generative AI use cases in fields like healthcare, finance, and marketing.

1. Identify at least three industry applications of generative AI using Azure AI Foundry catalog.
2. Summarize how generative AI models address specific business needs.
3. Present findings highlighting benefits and challenges.
4. Discuss Azure service features that enable these solutions.

Expected Outcome: Students will understand diverse generative AI applications on Azure across industries.

Analyze generated content for potential bias or misinformation using Azure AI Foundry portal.

1. Generate text and images using Azure OpenAI models.
2. Review outputs for biased language, stereotypes, or false information.
3. Document examples and suggest mitigation strategies.
4. Discuss Azure tools that support ethical AI development.

Expected Outcome: Students will recognize ethical risks in generative AI and ways to address them.

Use Azure Responsible AI resources to design ethical generative AI workflows.

1. Review Microsoft's Responsible AI principles and Azure tools like Fairlearn and InterpretML.
2. Identify points in generative AI development to apply fairness and transparency checks.
3. Create a checklist for ethical generative AI deployment.
4. Present the ethical workflow design.

Expected Outcome: Students will learn to incorporate responsible AI practices in generative AI projects on Azure.

Create, deploy, and manage generative AI models with Azure OpenAI service.

1. Provision Azure OpenAI resource and configure access.
2. Deploy a GPT or DALL-E model to production endpoint.
3. Use Azure portal and CLI to monitor usage and performance.
4. Update model deployment with new parameters or versions.

Expected Outcome: Students will gain practical skills in managing generative AI models on Azure.

Browse Azure AI Foundry catalog and integrate a selected generative AI model into an application.

1. Access Azure AI Foundry catalog and review available generative AI models.
2. Select a model suitable for content creation or code generation.
3. Follow integration instructions to connect the model to a sample application.
4. Test the application and document integration steps.

Expected Outcome: Students will understand Azure AI Foundry capabilities and model integration processes.

This example demonstrates how to identify the correct AI workload type based on a given scenario.

Answer / Conclusion: Scenario 1: NLP workload Scenario 2: Document Processing workload Scenario 3: Generative AI workload

• Understand characteristics of AI workload types
• Apply AI workload knowledge to real-world use cases

This example helps learners distinguish between different computer vision workloads such as image classification, object detection, and facial recognition.

Answer / Conclusion: Use Case A: Image Classification Use Case B: Object Detection Use Case C: Facial Recognition

• Differentiate between computer vision workload types
• Recognize Azure services supporting these workloads

Demonstrates how to identify NLP tasks such as sentiment analysis, entity recognition, and translation, and associate them with Azure AI Language services.

Answer / Conclusion: Task 1: Sentiment analysis using Azure AI Language Task 2: Named Entity Recognition using Azure AI Language Task 3: Translation using Azure Translator service

• Identify common NLP workloads
• Match NLP tasks to Azure AI Language features

Illustrates how OCR and form processing work within document processing workloads on Azure.

Answer / Conclusion: Use Azure Document Intelligence for OCR and form processing to extract data from invoices.

• Understand OCR and form processing concepts
• Identify Azure services for document processing

Explains how to determine when to use generative AI models for content creation, code generation, and chatbots.

Answer / Conclusion: Scenario 1: Content creation Scenario 2: Code generation Scenario 3: Chatbots

• Recognize generative AI use cases
• Understand Azure services supporting generative AI

Describes steps to identify bias and apply mitigation techniques using Azure tools.

Answer / Conclusion: Bias detected via fairness metrics; mitigated using Azure Responsible AI tools and techniques

• Understand bias detection methods
• Learn mitigation strategies and Azure tool support

Explains how to ensure AI model reliability and safety using Azure monitoring tools.

Answer / Conclusion: Reliable, safe AI solutions maintained via Azure ML monitoring and retraining pipelines

• Importance of monitoring AI models
• Azure tools for reliability and safety

Guides on implementing privacy and security controls for AI workloads on Azure, including GDPR compliance and data anonymization.

Answer / Conclusion: GDPR compliant and secure AI system using Azure privacy and security features

• Understand GDPR and privacy principles
• Apply Azure security best practices

Shows how to enhance AI model transparency and interpretability using Azure tools.

Answer / Conclusion: Transparent AI models with clear explanations using Azure Interpretability tools

• Techniques for AI transparency
• Azure tools supporting explainability

Illustrates how to implement governance frameworks and audit trails for AI solutions using Azure features.

Answer / Conclusion: Accountable AI systems governed and audited using Azure governance tools

• Understand AI accountability principles
• Implement Azure governance and audit features

This example shows how to identify regression problems such as forecasting and price prediction and the appropriate Azure tools.

Answer / Conclusion: Regression problems identified and matched with Azure ML regression tools

• Recognize regression problem characteristics
• Understand Azure support for regression modeling

Demonstrates identifying classification problems including binary and multi-class classification and corresponding Azure services.

Answer / Conclusion: Classification problems recognized and matched with Azure classification services

• Differentiate binary vs. multi-class classification
• Use Azure tools for classification scenarios

Explains clustering as unsupervised learning and typical use cases like customer segmentation and anomaly detection with Azure support.

Answer / Conclusion: Clustering scenarios identified with Azure ML support

• Understand clustering and unsupervised learning
• Recognize Azure tools for clustering

Clarifies the difference between features (inputs) and labels (outputs) in datasets used for training ML models.

Answer / Conclusion: Features: Age, Income, Credit Score; Label: Loan Approval Status

• Understand role of features and labels
• Prepare datasets correctly for ML training

Describes common deep learning architectures and their applications, highlighting Azure ML support for training such models.

Answer / Conclusion: Deep learning CNN architectures applied using Azure ML GPU resources

• Know deep learning architectures and use cases
• Utilize Azure ML for deep learning

Introduces the Transformer architecture and the attention mechanism used in language modeling, with Azure AI services using Transformers.

Answer / Conclusion: Transformer architecture powers Azure NLP services

• Understand Transformer and attention concepts
• Recognize Azure use of Transformer models

Demonstrates categorization of computer vision workloads such as image classification, object detection, OCR, and video indexing.

Answer / Conclusion: Tasks categorized into computer vision workload types

• Identify common computer vision workload categories
• Link tasks to Azure computer vision services

Walks through the process of preparing data and training an image classification model using Azure Custom Vision service.

Answer / Conclusion: Successfully trained and deployed image classification model

• Understand image classification workflow
• Leverage Azure Custom Vision capabilities

Explains how to prepare data and train an object detection model, highlighting bounding box annotations and Azure Custom Vision features.

Answer / Conclusion: Object detection model trained and ready for deployment

• Learn object detection training process
• Use bounding box annotations effectively

Demonstrates the process of extracting text from scanned documents using Azure Computer Vision OCR capabilities.

Answer / Conclusion: Text content extracted from scanned handwritten letter

• Use Azure OCR for text extraction
• Understand OCR output format

Explains how to use Azure Video Indexer to analyze video content for transcription, face detection, and key moment identification.

Answer / Conclusion: Video transcription and face detection results available

• Leverage Azure Video Indexer features
• Understand video content analysis capabilities

Categorizes NLP workloads such as sentiment analysis, key phrase extraction, and speech recognition relevant to Azure AI Language and Speech services.

Answer / Conclusion: NLP tasks categorized and mapped to Azure services

• Identify common NLP workload types
• Understand Azure AI Language and Speech capabilities

Explains how to interpret sentiment analysis results, including score ranges and classification.

Answer / Conclusion: Sentiment classified as Positive with high confidence

• Understand sentiment score interpretation
• Apply sentiment analysis results in applications

Shows how to extract key phrases from text using Azure AI Language and apply the information.

Answer / Conclusion: Key phrases such as 'product quality', 'fast delivery' extracted

• Extract important phrases for text summarization
• Use Azure AI Language key phrase extraction feature

Demonstrates detecting the language of input text using Azure AI Language service for multi-language processing.

Answer / Conclusion: Detected language: French (fr)

• Use language detection for multi-language applications
• Leverage Azure AI Language language detection feature

Shows how to identify entities such as organizations, locations, and people in text using Azure AI Language NER.

Answer / Conclusion: Named entities: Microsoft (Organization), Redmond (Location), Bill Gates (Person)

• Understand named entity recognition
• Apply Azure AI Language NER capabilities

Demonstrates translating text from one language to another using Azure Translator service.

Answer / Conclusion: Translated text: 'Hola, ¿cómo estás?'

• Use Azure Translator for language translation
• Integrate translation in AI applications

Describes capabilities and limitations of generative AI models like GPT-4o and DALL-E available via Azure OpenAI service.

Answer / Conclusion: Overview of generative AI model types and Azure OpenAI service features

• Understand generative AI model types
• Recognize Azure OpenAI service capabilities

Highlights common generative AI use cases such as content creation, virtual assistants, and code generation with industry examples.

Answer / Conclusion: Use cases mapped to industries using Azure generative AI solutions

• Identify generative AI applications by industry
• Leverage Azure AI Foundry models

Discusses ethical risks in generative AI such as bias and misinformation, and tools on Azure to address them.

Answer / Conclusion: Ethical generative AI deployment with bias mitigation and monitoring

• Recognize ethical risks in generative AI
• Apply Azure Responsible AI tools for mitigation

Explains how to use Azure OpenAI service and AI Foundry for deploying and managing generative AI models.

Answer / Conclusion: Generative AI chatbot deployed and managed using Azure services

• Utilize Azure OpenAI for generative AI models
• Manage models with Azure AI Foundry

This concept map illustrates the main AI workload categories available on Azure including computer vision, NLP, document processing, and generative AI, with examples of use cases for each workload type.

mermaidgraph TD
  A[Azure AI Workloads] --> B[Computer Vision]
  A --> C[Natural Language Processing]
  A --> D[Document Processing]
  A --> E[Generative AI]
  B --> B1[Image Classification]
  B --> B2[Object Detection]
  B --> B3[Facial Recognition]
  C --> C1[Sentiment Analysis]
  C --> C2[Entity Recognition]
  C --> C3[Translation]
  D --> D1[OCR]
  D --> D2[Form Processing]
  E --> E1[Content Creation]
  E --> E2[Code Generation]
  E --> E3[Chatbots]

Usage: Use this visual early in the lesson to introduce and differentiate the main AI workload categories on Azure.

This diagram shows the components of a typical computer vision workload on Azure, including image input, Azure Custom Vision service, model training and deployment, and use cases like object detection and facial recognition.

mermaidgraph TD
  A[Image/Video Input] --> B[Azure Custom Vision Service]
  B --> C[Model Training]
  B --> D[Model Deployment]
  D --> E[Image Classification]
  D --> F[Object Detection]
  D --> G[Facial Recognition]
  D --> H[Video Processing]

Usage: Present this diagram when explaining how computer vision workloads are structured and deployed on Azure.

This flowchart depicts the sequence of processing in Azure NLP workloads, from text input through key NLP tasks like sentiment analysis, entity recognition, and translation, highlighting Azure AI Language and Speech services.

mermaidgraph TD
  A[Text or Speech Input] --> B[Azure AI Language Service]
  B --> C[Sentiment Analysis]
  B --> D[Key Phrase Extraction]
  B --> E[Entity Recognition]
  B --> F[Language Detection]
  B --> G[Translation]
  A --> H[Azure Speech Service]
  H --> I[Speech Recognition]
  H --> J[Speech Synthesis]

Usage: Use this visual when covering the workflow and capabilities of Azure NLP and Speech services.

Sequence diagram illustrating document processing steps using Azure Document Intelligence and OCR services: document upload, OCR extraction, form data analysis, and output of structured data.

mermaidsequenceDiagram
  participant User
  participant AzureFormRecognizer
  User->>AzureFormRecognizer: Upload Document/Image
  AzureFormRecognizer->>AzureFormRecognizer: Perform OCR
  AzureFormRecognizer->>AzureFormRecognizer: Extract Form Data
  AzureFormRecognizer-->>User: Return Structured Data

Usage: Show this visual when explaining how document processing workloads operate on Azure.

Table comparing common generative AI model types such as GPT-4o and DALL-E, their use cases, and corresponding Azure services supporting these workloads.

mermaidclassDiagram
  class GenerativeAIModels {
    <<table>>
    +Model Type
    +Use Cases
    +Azure Service
  }
  GenerativeAIModels : GPT --> Content Creation
  GenerativeAIModels : DALL-E --> Image Generation
  GenerativeAIModels : GPT-4o --> Code Generation
  GenerativeAIModels : AzureOpenAIService --> Supports all above models

Usage: Use this visual to summarize generative AI model types and their Azure service support during the lesson.

Flowchart outlining the process of detecting and mitigating bias in AI models using Azure tools, from data collection through bias analysis to model adjustment and monitoring.

mermaidgraph TD
  A[Collect Training Data] --> B[Bias Detection with Azure Fairness Toolkit]
  B --> C[Analyze Bias Sources]
  C --> D[Mitigate Bias (Data/Model Adjustments)]
  D --> E[Validate Model Fairness]
  E --> F[Deploy & Monitor Model]
  F --> B

Usage: Present this visual when discussing fairness and bias management in AI models.

Diagram showing components involved in reliability and safety of AI solutions on Azure including model validation, monitoring, alerting, and continuous improvement loops.

mermaidgraph TD
  A[AI Model] --> B[Validation & Testing]
  B --> C[Deployment]
  C --> D[Azure Monitor & Alerts]
  D --> E[Anomaly Detection]
  E --> F[Feedback Loop to Model Updates]
  F --> B

Usage: Use this visual to explain how Azure supports ongoing reliability and safety of AI models.

Concept map depicting key privacy and security principles including GDPR compliance, data anonymization, encryption, and Azure security services supporting AI workloads.

mermaidgraph TD
  A[Privacy & Security] --> B[GDPR Compliance]
  A --> C[Data Anonymization]
  A --> D[Data Encryption]
  A --> E[Azure Security Features]
  E --> E1[Azure Key Vault]
  E --> E2[Microsoft Defender for Cloud]
  E --> E3[Role-Based Access Control]
  E --> E4[Network Security]

Usage: Show this visual when covering data privacy and security considerations in Azure AI solutions.

Flowchart illustrating typical regression problem workflow: data input, feature selection, model training, evaluation, and prediction using Azure ML services.

mermaidgraph TD
  A[Data Collection] --> B[Feature Selection]
  B --> C[Train Regression Model]
  C --> D[Evaluate Model]
  D --> E[Deploy Model]
  E --> F[Make Predictions]

Usage: Use this visual when explaining regression ML workflows and Azure support.

Sequence diagram showing the lifecycle of a classification ML model on Azure including data preprocessing, training, validation, deployment, and inference.

mermaidsequenceDiagram
  participant DataScientist
  participant AzureML
  DataScientist->>AzureML: Prepare Dataset
  AzureML->>AzureML: Train Classification Model
  AzureML->>AzureML: Validate Model
  AzureML->>AzureML: Deploy Model
  DataScientist->>AzureML: Request Predictions
  AzureML-->>DataScientist: Return Classification Results

Usage: Present this visual during classification ML lesson to clarify model lifecycle.

Flowchart describing the unsupervised learning process for clustering including data input, feature extraction, clustering algorithm application, and cluster analysis on Azure.

mermaidgraph TD
  A[Data Input] --> B[Feature Extraction]
  B --> C[Apply Clustering Algorithm]
  C --> D[Analyze Clusters]
  D --> E[Use Clusters for Insights]

Usage: Use this visual to explain clustering and unsupervised learning concepts.

Concept map illustrating dataset components including features and labels, differences between training and validation datasets, and basic feature engineering steps.

mermaidgraph TD
  A[ML Dataset] --> B[Features]
  A --> C[Labels]
  A --> D[Training Dataset]
  A --> E[Validation Dataset]
  B --> F[Feature Engineering]
  F --> G[Data Cleaning]
  F --> H[Feature Scaling]
  F --> I[Feature Selection]

Usage: Present this visual when teaching dataset concepts and preparation.

Class diagram showing components of deep learning architectures including input layer, multiple hidden layers (neurons), output layer, and Azure ML support components.

mermaidclassDiagram
  class DeepLearningModel {
    +InputLayer
    +HiddenLayers
    +OutputLayer
    +ActivationFunctions
  }
  class AzureML {
    +GPU Support
    +Distributed Training
    +Model Management
  }
  DeepLearningModel <|-- AzureML

Usage: Use this visual to explain deep learning model structure and Azure capabilities.

Diagram showing the Transformer architecture components such as the input embedding, multi-head self-attention layers, feed-forward layers, and output layers, highlighting Azure AI service integration.

mermaidgraph TD
  A[Input Tokens] --> B[Input Embedding]
  B --> C[Multi-Head Self-Attention]
  C --> D[Feed-Forward Layer]
  D --> E[Output Layer]
  E --> F[Azure AI Services]

Usage: Present this visual when introducing Transformer models and their Azure applications.

Concept map categorizing computer vision workloads into image classification, object detection, OCR, and video indexing with Azure tools supporting each.

mermaidgraph TD
  A[Computer Vision Workloads] --> B[Image Classification]
  A --> C[Object Detection]
  A --> D[Optical Character Recognition (OCR)]
  A --> E[Video Indexing]
  B --> F[Azure Custom Vision]
  C --> F
  D --> G[Azure Document Intelligence & Computer Vision]
  E --> H[Azure Video Indexer]

Usage: Use this visual to introduce categories of computer vision workloads.

Flowchart detailing the steps for image classification using Azure Custom Vision: data collection, labeling, training, evaluation, and deployment.

mermaidgraph TD
  A[Collect Images] --> B[Label Images]
  B --> C[Train Model in Azure Custom Vision]
  C --> D[Evaluate Model]
  D --> E[Deploy Model]
  E --> F[Make Predictions]

Usage: Show this visual during explanation of image classification workloads.