The goal is to design a large scale optimization algorithm for machine learning tasks that can run on decentralized compute environments (such as peer-to-peer systems). Requires ability to program and do empirical analysis and also understand theoretical optimization models/literature.