Fliptop's primary business used to be contact record linkage. We built-out some substantial infrastructure, coding primarily in Scala, and using JVM-stack libraries. In the past year, we've moved beyond our data business into building predictive analytics applications for Sales and Marketing teams, so it was necessary to build a data science/machine learning stack. While there's plenty of machine-learning tools for the JVM, we felt that the Python ecosystem had many advantages for our use-case (which we'll discuss), and we think we integrated them in a clever way.