The success of machine learning is heavily dependent on the choice of data features on which the methods are applied. For that reason, much of the actual efforts in deploying algorithms go into the design of features that support effective machine learning. In this talk, I describe our efforts to expand the scope and ease the applicability of machine learning on complex, interconnected datasets. First, I outline our methods for graph representation learning. The methods specify deep graph neural functions that map nodes in a graph to points in a compact vector space, termed embeddings. Importantly, these graph neural methods are optimized to embed the graph such that performing algebraic operations in the learned embedding space reflects topology of the graph. We show how embeddings enable repurposing of drugs for new indications and the discovery of dozens of drug combinations that are safe in patients with considerably fewer unwanted side effects than today's treatments. Lastly, I describe our efforts in learning actionable representations that allow users of our models to ask what-if questions and receive predictions that are accurate and can be interpreted meaningfully.