Contents

A short description of each topic or subtopic.

• MLOps Theory: A brief introduction to the main Principles of MLOps: Data and Model Versioning, Feature Management and Storing, Automation of Pipelines and Processes, CI/CD for Machine Learning and Continuous Monitoring of Models. As well as Common Tools used to address each of those points.

• Implementations Guide

  • Introduction
    • Tools and Project Structure: Introduction to the project structure and tools that will be used.
    • Starting a New Project with Cookiecutter: Introduction to Cookiecutter and how to create a new project based on our template.

  • Environment

    • Setting Up the IBM Environment with Terraform: Using Terraform to set up the IBM Environment via code.
    • Managing the deployment space: How to manage IBM's tools with Terraform.

  • Versioning

    • What is DVC?: Introduction to DVC, installation and how to setup remote storage.
    • Data Versioning: Working with DVC to version data and models.
    • Working with Pipelines: Creating and reproducing pipelines for training and evaluating models.

  • Deployment

    • Deployment with Watson Machine Learning: Using Watson ML API to deploy models as an online web service.

  • CI/CD for Machine Learning

    • Continuous Integration with CML and Github Actions: Introduction to CML and GitHub Actions and how to create automatic testing and reportng workflows.
    • Continous Delivery with CML, Github Actions and Watson ML: Using CML and GitHub Actions create automatic deployments for every new release.

  • Monitoring

    • Monitoring with IBM OpenScale: Setting up OpenScale environment, creating monitors, evaluating the model and explaining predictions.

  • Project Workflow

    • Project Workflow: Demonstrating how to use the complete pipeline in a development cycle we created with commands and videos.