Efficient Model Exploring and Continuous Delivery With Polyaxon + Kubeflow

Explore efficient model development and continuous delivery using Polyaxon and Kubeflow in this 19-minute conference talk from KubeCon + CloudNativeCon North America 2021. Discover how Mercari's Machine Learning Platform team accelerates ML projects through the integration of these cloud-native tools. Learn about the iterative nature of machine learning projects, from experimentation to productionization and operation. Gain insights into Polyaxon's capabilities for parallel and scalable hyperparameter tuning, and understand how KubeflowPipelines serves as a workflow engine for ML pipelines. Delve into Mercari's ML development lifecycle, exploring their use of Polyaxon and Kubeflow Pipelines. Uncover the team's innovative solutions, including a monorepo for Kubeflow Pipelines, a "Project" manifest for KFP and Polyaxon, and a Polyaxon Kubeflow Pipelines component. Walk away with valuable takeaways to enhance your own ML project workflows and accelerate iterations.

Intro
What is Mercari?
Machine Learning at Mercari US
ML Development Lifecycle
ML Project Lifecycle at Mercari US
What is Polyaxon?
How to run a job on Polyaxon
What is Kubeflow Pipelines?
Kubeflow Pipelines at Mercari US
What we built to accelerate iterations
Monorepo for Kubeflow Pipelines
"Project" Manifest for KFP and Polyaxon
Polyaxon Kubeflow Pipelines Component
Takeaways