Species interactions underpin a lot of ecological processes, and yet they are notoriously difficult to sample. We develop techniques from applied machine learning to predict species interactions from functional traits and phylogenetic information, and build sampling algorithms to better understand where we should go look for new interactions.