ML on Google Cloud Whitepaper

Google Cloud

Learn about ML on Google Cloud in this whitepaper by Devoteam G Cloud AI & ML experts.

Reading this white paper, written by our cloud engineers, will help orient you in getting started in the world of Machine Learning and MLOps on Google Cloud.

Download the White paper to discover:

  • basic knowledge of Google Cloud
  • basic knowledge of MLOps
  • overview of machine learning services on Google Cloud
  • how to apply MLOps on Google Cloud

Table of contents

1 Introduction

2 The basics of Google Cloud

2.1 The GCP Resource Hierarchy

2.2 Identity and Access Management

2.3 Services

2.4 Locations, regions and zones

3 The basics of MLOps

3.1 Towards a reusable deployment pipeline

3.1.1 Pipeline components

3.1.2 Pipelines

3.1.3 Version Control System

3.1.4 Building and storing artifacts

3.1.5 An end to end example

3.2 The Feature Store

3.3 Environment management

3.4 Monitoring of deployed models

3.4.1 SLAs, SLOs and SLIs

3.4.2 Meeting SLOs

3.4.3 Measuring Model Performance SLIs

3.5 Monitoring for Model Drift

3.6 Keeping models up to date

3.7 Infrastructure as code

4 Machine Learning services on Google Cloud

4.1 Cloud Storage

4.2 BigQuery

4.3 Vertex AI

4.3.1 Vertex AI Training

4.3.2 Vertex AI Pipelines

4.3.3 Vertex AI Metadata Store

4.3.4 Vertex AI Model Registry

4.3.5 Vertex AI Feature Store

4.3.6 Vertex AI Workbench

5 Applying MLOps on Google Cloud

5.1 Data Exploration and Experimentation

5.2 Vertex AI Feature Store

5.3 Pipelines on Vertex AI

5.4 Model Training

5.4.1 AutoML Training

5.4.2 Custom Training

5.4.3 Custom Hyperparameter Tuning Training

5.5 Model Hosting

5.5.1 Vertex AI Endpoints

5.5.2 Cloud Run

5.5.3 Google Kubernetes Engine

5.5.4 When to choose which option?

5.6 Model Monitoring

5.7 Model transparency and fairness

5.8 A practical example

6 Conclusion