I'll describe algorithms and data structures for allowing the most powerful machine learning and multivariate statistical methods, which often scale quadratically or even cubically with the number of data points, to be performed many orders of magnitude faster than naive implementations.