Real-world data doesn’t come pre-packaged and ready to use neatly right out of the box. Its raw, messy, sparse, and exists across untold numbers of different formats. Unclean and poorly formatted data can instantly sacrifice the capabilities of even the best analytic objectives and machine learning models. This talk centers on best practices and methods for wrangling data, reformatting to make data easier to work with and more flexible. Additionally it focuses on methods for feature engineering--to derive from your raw data those representative elements and structures that represent exactly what you wish to test. Most projects are won or lost at the wrangling and feature engineering stage; the right tools can make all the difference.