Abstract: This tutorial will demonstrate how to use Kubeflow Pipelines to create a full Machine Learning application on Kubernetes.
Productionizing Machine Learning workloads is today one of the key challenges in turning Machine Learning models into reliable drivers of business value. Model training and retraining, model serving and integration with operational systems, model validation, data validation and data generation are the key components of a robust, production level Machine Learning application, or pipeline. This comes with an inherent complexity that Kubeflow Pipelines attempts to solve for and make it manageable for Engineers and Data Scientists alike.
Kubeflow is an Open Source curated set of compatible tools and artifacts that lays a foundation for running production Machine Learning applications. It enables consistency across deployments throughout all environments from development to production, by providing Kubernetes object templates that bring together disparate components. Kubeflow’s purpose is to make it easy for everyone to develop, deploy, and manage portable and scalable Machine Working workloads everywhere.
The tutorial will cover how to build and run a complete Machine Learning pipeline that does distributed training of a TensorFlow model. The trained model will further be deployed for serving predictions at scale and a Web frontend will be deployed to demonstrate requesting predictions from the trained model. In addition, participants will learn how to use a Jupyter notebook to build and run a Kubeflow pipeline using the Kubeflow Pipelines SDK.
The cloud platform on which the Kubeflow Pipelines Machine Learning application will be deployed in this tutorial is Google Cloud Platform. Participants will thus become familiar with Google Cloud Platform tools such as Cloud Shell, Kubernetes Engine and Deployment Manager.
Bio: Dan Anghel is a Machine Learning Cloud Engineer working at Google for over 4 years. Prior to Google, Dan has spent more than 10 years in the French retail industry building scalable E-commerce solutions. Specialized in Machine Learning and Big Data, he is helping the largest Google customers build scalable production level solutions on Google Cloud