Continuous intelligence: Moving machine learning into production reliably

Danilo Sato and Christoph Windheuser walk you through applying continuous delivery (CD), pioneered by ThoughtWorks, to data science and machine learning. Join in to learn how to make changes to your models while safely integrating and deploying them into production, using testing and automation techniques to release reliably at any time and with a high frequency.

So you want to include a machine learning component in your IT systems? The process is a little more involved than clicking through an AI tutorial on your laptop. It’s not just the first working model you run that you need to consider; you also need to think about things like integration, scaling, and testing. What’s more, postlaunch, you’ll want to continuously adapt your model to respond to the changing environment. ThoughtWorks pioneered continuous delivery—a set of tools and processes that ensure that software under development can be reliably released to production at any time and with high frequency. Danilo Sato and Christoph Windheuser demonstrate how to apply continuous delivery to machine learning—what’s known as continuous intelligence. In a live scenario, you’ll change a machine learning model in a development environment, test its new performance, and, depending on the outcome, automatically deploy the new model into a production environment. The tech stack for this scenario will be Python, DVC (Data Science Version Control), and GoCD.