One of the most popular, and useful, ways to productionize a machine learning solution is scheduled batch workflows. In this approach, we deliver predictions in regular intervals. There are many tools available allowing you to construct and schedule your workflows. When there are many options, it can be difficult to choose.
In this session, we will talk about how and why we switched from Kubeflow to Argo Workflows for batch workflows; how we approach employing a new MLOps tool; and why simplicity is the way to go forward.
One of the most popular, and useful, ways to productionize a machine learning solution is scheduled batch workflows. In this approach, we deliver predictions in regular intervals. There are many tools available allowing you to construct and schedule your workflows. When there are many options, it can be difficult to choose.
In this session, we will talk about how and why we switched from Kubeflow to Argo Workflows for batch workflows; how we approach employing a new MLOps tool; and why simplicity is the way to go forward.