< Back to BLOG /

Building Reproducible Automated ML Workflow with Kubeflow and Hydrosphere

Very often a workflow of training machine learning models and delivering them to production environment contains loads of manual work. These could be various steps depending on the type of a model you are using, company’s workflow you are working within and requirements of the deployed model.

Industry has already developed tools for continuous software delivery/integration, but they cannot be directly applied for machine learning models which designates the problem.

Read our article on Medium, where we describe a way to create a pipeline that connects machine learning workflow steps (like collecting & preparing data, model training, model deployment and so on) into a single reproducible run, which you can execute with a single button push.


We also looking forward to meet you on our workshop on this topic at ODSC East 2019, May 1st, Boston (MA).