Introduction to AI in Microsoft Azure AI-900 Labs

Labs for hands on work related to Introduction to AI in Azure - Training | Microsoft Learn with labs described at Microsoft Learn - Introduction to AI in Azure exercises | mslearn-ai-fundamentals

Each lab uses sign in at https://ai.azure.com and then manages resources at https://ai.azure.com/managementCenter/AllResources. Each lab takes 20-30 minutes though with exploration can take 1 hour.

Explore Automated Machine Learning in Azure Machine Learning (1 hour)

Explore Automated Machine Learning in Azure Machine Learning | mslearn-ai-fundamentals

Working on a model predicts number of bicycle rentals expected in a day, based on seasonal and meteorological features.

Create new Azure ML resource
Use the Studio at https://ml.azure.com
Authoring > Automated ML > New job with settings using
- Task type: Regression
- Primary metric: Normalized Root Mean Squared Error
- Wait for job to finish, can take a while
In Overview, view algorithm name and Metrics tab for residuals (difference in predicted and actual values) and predicted true chart comparing against true values.
Deploy and test the model: Model tab and select Deploy using Real-time endpoint
Test in ML studio > Endpoints > predict-rentals endpoint > Test tab
Use lab input data to test

Explore generative AI in Azure AI Foundry portal (1 hour)

Source: Explore generative AI in Azure AI Foundry Portal | mslearn-ai-fundamentals

Create new AI Foundry resource, project in existing RG
Playgrounds > Try chat > Deploy new gpt-4o model
Use chat history interface
Provide prompts about a Paris trip and iterate on locations, accommodations and city history

Analyze text in Azure AI Foundry portal (1 hour)

Analyze text in Azure AI Foundry portal | mslearn-ai-fundamentals

Create new AI Foundry resource, project in existing RG
Playground > Try the Language playground
Extract information > Extract named entities
Using a text sample below, run to review extracted entities

Tired hotel with poor service
The Royal Hotel, London, United Kingdom
5/6/2018
This is an old hotel (has been around since 1950's) and the room furnishings

Next, select Extract key phrases and run, note details of subjects
Summarize information tile > Summarize text, run on a text sample. Note the Extractive summary (use key phrases directly from text)

Explore speech in Azure AI Foundry portal (1 hour)

Explore speech in Azure AI Foundry portal | mslearn-ai-fundamentals

Create new project
Playgrounds > Try Speech Playground
Go to Real-time transcription
Upload lab audio, listen to recording. In output Text, see transcription

Analyze images in Azure AI Foundry portal (1 hour)

Analyze images in Azure AI Foundry portal | mslearn-ai-fundamentals

Create new project > AI services > Vision + Document
Go to Image > Image captioning tile
Add captions to images page > Upload lab images
See generated caption text, repeat for Dense captioning (descriptions for image’s content and other objects detected)
Hover over an attribute in Detected attributed list
Return to Vision + Document > Image > Common tag extraction tile
Upload lab image and note tags and confidence scores
Return to Vision + Document > Image > Common object detection
Upload lab image, see Detected attributes of objects and scores, adjust Threshold slider to 70 to see changes

Extract data with Content Understanding in Azure AI Foundry portal (1 hour)

Extract data with Content Understanding in Azure AI Foundry portal | mslearn-ai-fundamentals

Create new project in the Azure AI services > Try Content Understanding tile
Create new task for understand an invoice
Upload lab invoice file in Define Schema and use the Invoice data extraction template, complete analysis
In Test Analyzer, review fields and confidence scores, check JSON results
Build an analyzer: Analyzer list > + Build Analyzer with invoice details
Test the built analyzer using lab file, look at code and endpoint variable in code