Increasingly, companies are adopting tools, processes, and organizational structures that enable data scientists to focus most of their time and energy on model development. In this post, Ankur Kalra and Kelly Davis describe how and why CNN implemented Metaflow to enable their ML researchers to iterate faster and more effectively. Specifically, Metaflow eliminates the need for ML researchers to rewrite code and/or manage workflow scheduling and compute resources when transitioning from local development to cloud deployment and/or scaling up. They also discuss why they contributed a Terraform stack for deploying Metaflow.