1. Home
  2. DevOps for AI and Machine Learning

DevOps for AI and Machine Learning

MLOps is the application of DevOps practices and principles to the machine learning (ML) process. In DevOps for AI and Machine Learning, DevOps engineers and testers will learn the foundational knowledge and practical skills necessary to integrate machine learning operations (MLOps) into existing AI Model development workflows.

This workshop covers the entire MLOps lifecycle, focusing on essential concepts, tools, and methodologies required to deploy and maintain machine learning models within DevOps environments.

Key takeaways from this class include:

  • Gaining an in-depth understanding of MLOps principles and their integration with DevOps practices.
  • Learning to set up automated data engineering pipelines.
  • Tracking model experiments effectively.
  • Implementing CI/CD pipelines for machine learning models.
  • Establishing robust monitoring and retraining strategies.
  • Navigating security and compliance considerations in MLOps.

Throughout this workshop, students will gain real-world context through practical hands-on exercises, such as setting up feature stores, implementing CI/CD processes for ML models, and deploying and monitoring models.

Who Should Attend
This workshop is ideal for DevOps engineers, software testers, and operations personnel looking to expand their skill set into MLOps. Professionals involved in software development, deployment, infrastructure management, quality assurance, or operations who wish to understand the unique challenges and best practices in deploying and maintaining machine learning models will benefit. It caters to individuals with a technical background in DevOps practices but with limited exposure to machine learning, aiming to bridge the gap between traditional DevOps workflows and the specialized requirements of MLOps.

Course Outline

Session 1: Introduction to AI/ML and MLOps

  • What is Artificial Intelligence (AI)?
  • Concerns in AI and Machine Learning (ML)
  • Steps and challenges in the ML process
  • Definition and importance of MLOps
  • Key MLOps activities
  • Example AI-based application
  • Exercise #1: Review example application

Session 2: MLOps During Model Development

  • What is model development?
  • Model experimentation process
  • Exercise #2: Using Jupyter notebooks
  • Creation of training datasets
  • Model experimentation
  • Experiment tracking
  • Exercise #3: Experiment track with mlflow
  • Model training pipelines
  • Exercise #4: Using nbconvert to export notebook code

Session 3: MLOps During Model Testing and Deployment

  • Approaches to testing and deploying AI models
  • Types of testing to perform
  • Model management
  • Using a model registry for model tracking
  • Exercise #5: Using mlflow to register and track models
  • Continuous delivery / deployment process
  • Exercise #6: Test and deploy models using mlflow and pytest
  • Online experimentation
  • Integration of AI models into applications

MLOps During Model Inference and Monitoring

  • Prediction serving process
  • Model monitoring
  • Types of model monitoring
  • Dealing with model decay
  • Exercise #7: Identify types of model decay
  • Model retraining

Session 5: MLOps for Dataset and Feature Engineering

  • What are features?
  • Defining and managing features
  • Dataset management
  • Using feature stores
  • Exercise #8: Identifying steps in the dataset and feature management process

Session 6: Model Governance and Compliance

  • Model governance vs. compliance
  • Types of model governance
  • Explainability
  • Fairness and bias
  • Data security and privacy
  • Model compliance standards
  • Exercise #9: Identifying types of governance

Putting It All Together and Next Steps

  • Comprehensive ML workflow
  • Summary and wrap-up of the course.
  • References
  • Q&A session to address participant queries.