I.  Overview:

This course introduces participants to MLOps tools and best practices for deploying, evaluating, monitoring and operating production ML systems on Google Cloud. MLOps is a discipline focused on the deployment, testing, monitoring, and automation of ML systems in production. Machine Learning Engineering professionals use tools for continuous improvement and evaluation of deployed models. They work with (or can be) Data Scientists, who develop models, to enable velocity and rigor in deploying the best performing models.

II. Duration: 01 day

III. Objectives

• Identify and use core technologies required to support effective MLOps.
• Configure and provision Google Cloud architectures for reliable and effective MLOps environments.
• Implement reliable and repeatable training and inference workflows.
• Adopt the best CI/CD practices in the context of ML systems.
• Operate deployed machine learning models effectively and efficiently.
• Integrate ML workflows with upstream and downstream data management workflows to maintain end-to-end lineage and metadata management.

IV. Intended Audience

• Data Scientists looking to quickly go from machine learning prototype to production to deliver business impact.
• Software Engineers looking to develop Machine Learning Engineering skills.
• ML Engineers who want to adopt Google Cloud.

V.    Prerequisites

• Completed Machine Learning with Google Cloud or have equivalent experience

VI.  Outline

The course includes presentations, demonstrations, and hands-on labs.

Module 1: Why and When do we need MLOps

• Discuss Data Scientists' pain points.
• Identify ML Engineering characteristics and challenges.
• Define how Google Cloud can help with MLOps.
• Recognize how MLOps differs from manual ML management.
• Compare and contrast DevOps vs MLOps.

Module 2: Understanding the Main Kubernetes Components (Optional)

• Define what is a Docker container.
• Create Docker containers.
• Identify the architecture of Kubernetes: pods, namespaces.
• Create Docker containers using Google Container Builder.
• Store container images in Google Container Registry.
• Create a Kubernetes Engine cluster.
• Manage Kubernetes deployments.

Module 3: Introduction to AI Platform Pipelines

• Identify the benefits and opportunities of AI Pipelines.
• Define Access Controls within AI Pipelines.
• Recognize pipeline components.
• List pipeline workflows.
• Set up AI Platform Pipelines.
• Create a machine learning pipeline.
• Run a machine learning pipeline.
• Connect to AI Platform Pipelines using the Kubeflow Pipelines SDK.
• Configure a Google Kubernetes Engine cluster for AI Platform Pipelines.

Module 4: Training, Tuning and Serving on AI Platform

• Identify the main concepts of MLOps on AI Platform.
• Create a reproducible dataset.
• Implement a tunable model.
• Build and push a training container.
• Train and tune a model.
• Serve and query a model.

Module 5: Kubeflow Pipelines on AI Platform

• Recognize how Kubeflow Pipelines fits in MLOps.
• Describe a Kubeflow Pipeline with KF DSL.
• Use the various Kubeflow components.
• Compile, upload, and run a pipeline build in Kubeflow Pipelines.

Module 6: CI/CD for Kubeflow Pipelines on AI Platform

• Create Cloud Build Builders.
• Configure pipelines with Cloud Build.
• Create triggers for training models using Cloud Build Triggers.
• Adopt the best CI/CD practices in the context of ML systems.

Module 7: Summary

Summarize the course.