Machine learning application development is becoming popular nowadays. The process of developing machine learning applications is often more complex compared to traditional software development. This is due to the nature of machine learning, where behavior is often hard to predict, hard to reproduce, thus hard to test and therefore difficult to bring into production. How does one setup and automate a pipeline and incorporate a workflow that is suitable for machine learning?

During this presentation we will walk through on how we setup a continuous integration & continuous deployment pipeline and how we test, train and deploy machine learning applications. We also show how we incorporate a workflow that let data scientists focus on machine learning rather than infrastructure.