This presentation will explore the application of extreme value theory and tail-dependence modelling to portfolio allocation, beginning with a simple two-asset case and then moving on to a more realistic multi-asset case that uses results from machine learning to allow the optimisation calculations to run in a reasonable time.