With the ability to scale in minutes we have been able to test various approaches and deployment strategies to get the best performing model for our particular problem space. I would definitely recommend this framework to any of my colleagues who are looking for an easy way to get started with machine learning. One thing that was great was having everything automated through scripts (pipeline). As soon as you put your data into S3 all the necessary steps happen automatically so there really isn't anything manual at all! Also one feature that we liked was being able to choose different ML algorithms from scikit-learn which were available when using the pipeline API.