Responsible AI workshop

​This session equips you with hands-on skills to build reliable and ethical AI models using Azure tools.

Friday Day 1: 0830-1430hrs
Saturday Day 2: 0830-1430hrs

​What you'll learn:

• ​Debug & improve models: Use the Responsible AI Dashboard to analyze data, understand errors, and ensure fairness (all while avoiding harmful content).

• ​Detect & block harmful content: Engineer prompts for AI models to identify and block unwanted text and images (think hate speech, violence, etc.).

• ​Create & evaluate AI workflows: Build and visualize workflows that combine language models, prompts, and tools to design and test your AI applications effectively.

​A simplified breakdown of each lab:

​1. **Responsible AI Dashboard**:

​ - **Objective**: This lab focuses on understanding and addressing issues that may arise when using machine learning models.

​ - **Topics Covered**:

​ - **Error Analysis**: Identifying and understanding errors made by the model.

​ - **Data Analysis**: Analyzing the quality and characteristics of the data used to train the model.

​ - **Model Explainability**: Understanding how the model makes predictions and explaining its decisions.

​ - **Model Performance**: Evaluating how well the model performs on different tasks.

​ - **Fairness Assessment**: Assessing whether the model behaves fairly across different demographic groups.

​ - **Hands-on Activities**: Participants will learn how to identify and address issues in machine learning models using various techniques, such as analyzing data, interpreting model predictions, and evaluating model fairness.

​2. **Azure Content Safety for Azure OpenAI**:

​ - **Objective**: This lab focuses on ensuring that AI models are used responsibly by detecting and mitigating harmful content.

​ - **Topics Covered**:

​ - **Prompt Engineering**: Designing prompts or inputs to AI models to achieve specific goals.

​ - **Detecting Harmful Content**: Identifying offensive or inappropriate content in text and images.

​ - **Mitigating Content Harm**: Implementing measures to reduce the impact of harmful content, such as sexual, violent, hate speech, and self-harm content.

​ - **Hands-on Activities**: Participants will learn how to engineer prompts, detect harmful content using AI models, and implement strategies to mitigate the impact of such content.

​3. **Azure Machine Learning Prompt Flow**:

​ - **Objective**: This lab focuses on creating and evaluating workflows that leverage language models and other tools for AI applications.

​ - **Topics Covered**:

​ - **Creating Executable Flows**: Designing workflows that incorporate language models, vector embeddings, prompts, and Python tools.

​ - **Visualizing Workflows**: Representing workflows graphically to understand the flow of data and processes.

​ - **Evaluating Performance**: Assessing the performance of AI applications through large-scale testing and analysis.

​ - **Hands-on Activities**: Participants will learn how to create, visualize, and evaluate workflows for AI applications using Azure Machine Learning tools.

​These labs provide participants with practical experience and skills to effectively use AI technologies responsibly, detect and mitigate harmful content, and create and evaluate AI applications.

​​Prerequisites :

​​- Basic knowledge of python programming.

- If you don't have experience with Python prerequisite material has been made available that you need to complete before the 29th that covers introduction to Python, ML and Cloud Computing

​​- GitHub account: if you don't have a github account, signup using this link: github.com/signup

​​- Microsoft Azure account: You can create your azure account using this link: azure.microsoft.com/free/