DP-100 Designing and Implementing a Data Science Solution on Azure Training

Module 1: Explore Azure Machine Learning Workspace Resources and Assets

1. Create an Azure Machine Learning workspace

2. Identify workspace resources and their purposes

3. Explore Azure Machine Learning assets (datasets, models, environments)

4. Train and manage models within the workspace

5. Understand asset tracking and versioning

Module 2: Explore Developer Tools for Workspace Interaction

1. Use Azure Machine Learning Studio for no-code interaction

2. Access and manage resources using the Python SDK

3. Automate tasks using the Azure CLI

4. Understand tool interoperability within the workspace

5. Choose the right tool based on use case

Module 3: Make Data Available in Azure Machine Learning

1. Understand the role of URIs in data access

2. Create and configure datastores for data storage

3. Create reusable data assets for experiments

4. Manage dataset versioning and access control

5. Organize data for scalable ML workflows

Module 4: Work with Compute Targets in Azure Machine Learning

1. Choose between compute instance, cluster, or attached compute

2. Create and use a development compute instance

3. Scale workloads using compute clusters

4. Optimize resource allocation for training jobs

5. Monitor compute usage and performance

Module 5: Work with Environments in Azure Machine Learning

1. Understand Azure ML environments and dependencies

2. Use curated environments for rapid experimentation

3. Create custom environments with Conda or Docker

4. Register and reuse environments across jobs

5. Manage environment versions for reproducibility

Module 6: Find the Best Classification Model with Automated Machine Learning

1. Preprocess data and configure featurization settings

2. Launch an AutoML classification experiment

3. Track progress and results through the UI and SDK

4. Evaluate models using built-in metrics

5. Select the best-performing model for deployment

Module 7: Track Model Training in Jupyter Notebooks with MLflow

1. Configure MLflow for local and remote tracking

2. Train models in notebooks using MLflow APIs

3. Log parameters, metrics, and artifacts

4. View experiment results and history

5. Use MLflow UI for model comparison

Module 8: Run a Training Script as a Command Job in Azure Machine Learning

1. Convert a Jupyter notebook to a Python script

2. Submit a script as a command job

3. Pass parameters and inputs to the training job

4. Monitor job execution and outputs

5. Use scripts for repeatable and scalable ML runs

Module 9: Track Model Training with MLflow in Jobs

1. Log metrics and artifacts using MLflow

2. View job metrics in Azure Machine Learning UI

3. Evaluate model performance across runs

4. Compare jobs and select optimal models

5. Integrate MLflow with sweep jobs and pipelines

Module 10: Perform Hyperparameter Tuning with Azure Machine Learning

1. Define hyperparameter search space

2. Choose a sampling strategy (random, grid, Bayesian)

3. Set early termination policies for efficiency

4. Launch sweep jobs to optimize model parameters

5. Analyze tuning results and select best model

Module 11: Run Pipelines in Azure Machine Learning

1. Create reusable pipeline components

2. Chain components together into a pipeline

3. Submit pipeline jobs for orchestration

4. Manage pipeline versions and dependencies

5. Monitor pipeline runs and outputs

Module 12: Register an MLflow Model in Azure Machine Learning

1. Log trained models with MLflow

2. Understand the MLflow model format and structure

3. Register a model in the Azure Machine Learning registry

4. Track model versions and metadata

5. Prepare models for deployment and monitoring

Module 13: Create and Explore the Responsible AI Dashboard

1. Understand the importance of Responsible AI

2. Create the Responsible AI dashboard for a trained model

3. Evaluate model fairness, explainability, and error analysis

4. Interpret dashboard insights for decision-making

5. Use the dashboard to improve model transparency

Module 14: Deploy a Model to a Managed Online Endpoint

1. Understand managed online endpoint architecture

2. Deploy an MLflow model to an online endpoint

3. Deploy custom models to endpoints

4. Test and validate endpoint functionality

5. Monitor endpoint status and usage

Module 15: Deploy a Model to a Batch Endpoint

1. Understand batch inference and use cases

2. Create and configure batch endpoints

3. Deploy MLflow or custom models to batch endpoints

4. Invoke batch jobs and analyze output

5. Troubleshoot batch deployment issues

Module 16: Introduction to Azure AI Foundry

1. Define what Azure AI Foundry is

2. Explore how Azure AI Foundry integrates with ML workflows

3. Understand AI Foundry’s capabilities in managing models

4. Identify when and why to use AI Foundry

5. Compare AI Foundry to other model management tools

Module 17: Explore and Deploy Models from the Model Catalog

1. Navigate the Azure AI Foundry model catalog

2. Explore available foundation and language models

3. Deploy catalog models to endpoints

4. Enhance model performance using tuning options

5. Monitor usage and outcomes of deployed models

Module 18: Get Started with Prompt Flow

1. Understand the development lifecycle for LLM apps

2. Explore core components and flow types

3. Connect to data sources using prompt flow connections

4. Configure runtimes and variants

5. Monitor and iterate on flow performance

Module 19: Build a RAG-Based Agent with Your Own Data

1. Understand Retrieval-Augmented Generation (RAG) principles

2. Prepare and index your data for searchability

3. Ground a language model using your own data

4. Build an agent using prompt flow in Azure AI Foundry

5. Evaluate grounded model outputs for accuracy

Module 20: Fine-Tune a Language Model

1. Identify when fine-tuning is appropriate

2. Prepare training data for fine-tuning chat models

3. Use Azure AI Studio to configure fine-tuning jobs

4. Deploy fine-tuned models to endpoints

5. Monitor fine-tuning metrics and results

Module 21: Evaluate the Performance of Generative AI Apps

1. Set benchmarks for model evaluation

2. Perform manual evaluations of model output

3. Use metrics to assess model quality and responsiveness

4. Compare app iterations to improve performance

5. Integrate evaluation into the development cycle

Module 22: Responsible Generative AI

1. Plan a responsible generative AI solution lifecycle

2. Identify and document potential harms (bias, toxicity, etc.)

3. Define harm measurement strategies

4. Apply mitigation techniques in model training and deployment

5. Implement governance and operations for responsible AI