Introduction to Kubeflow

What you'll learn

• 
  Basic Kubeflow architecture and components
• 
  How to deploy Kubeflow on Google Kubernetes Engine
• 
  How to develop a Machine Learning model using Kubeflow Notebook
• 
  How to train and deploy a TensorFlow model in Kubeflow
• 
  How to productionise a TensorFlow model using KFServing
• 
  How to manage logs and meta information

Description

Kubeflow is an open-source Machine Learning platform built on top of Kubernetes. Kubeflow provides a toolkit that can be used for building and orchestrating Machine Learning pipelines.

In this mini-course Luis will teach you how to use Kubeflow in a practical manner. By the end of the course you will understand how Kubeflow works and you'll be able to start developing your own pipelines.

This course includes 5 short lectures, quizzes as well as source code for each screencast!

You will learn:

• Core Kubeflow components
• How to set up Kubeflow on Kubernetes
• How to develop basic ML models in Kubeflow Notebooks
• How to train and deploy models in Kubeflow
• How to use Kubeflow Pipelines
• How to use KFServing to deploy models
• How to manage logs with Kubeflow Metadata component

This course is for intermediate developers. We assume that you already have practical knowledge about Kubernetes, Docker and Google Cloud. You do not need any previous knowledge of Kubeflow.

Introduction video

Course content

5 lectures, 1h total length

Introduction

Course introduction (5:31) Preview

In this lecture Luis will give you a brief overview of the course curriculum and Kubeflow (ML toolkit of Kubernetes).

Infrastructure Deployment

Deploying Kubeflow on Google Cloud (7:24)

You're going to learn how to deploy Kubeflow v1.1 on Google Cloud Platform. You will make use of a Google Kubernetes Engine and Cloud Config Connectors.

Model Development

Model development (23:20)

You'll learn how to develop a Machine Learning model in Kubeflow using a Notebook. You'll also make use of training operators and Kubeflow pipelines.

Model Deployment

Model deployment (6:25)

You will use KFServing to deploy a Tensorflow model in your Kubeflow environment.

Metadata Management

Metadata management & course wrap-up (5:53)

You'll learn about Kubeflow's metadata component which can be used for tracking logs and meta information.

Who this course is for

• Machine Learning Engineers
• Data Engineers
• Data Scientists
• Python Developers Interested in MLOps

Requirements

• Intermediate Python programming knowledge
• Intermediate Docker knowledge
• Intermediate practical Kubernetes knowledge
• Familiarity with Google Cloud and having a Google Cloud account

Instructor

Luis Velasco
Machine Learning and Data Engineer @ Google

Luis is a Data Analytics and Engineering specialist at Google and a passionate content creator.

He is passionate about democratising Data Analytics and Machine Learning knowledge. He regularly publishes notebooks and blog posts in Data Science, DataOps and MLOps topics. Luis is familiar with each and every step of the data pipeline although he is most interested in enterprise scale Machine Learning deployments.

Luis also has a YouTube channel where he regularly creates videos about MLOps use cases in Spanish.