I mostly work on characterizing and improving the learning of large data sets, by systematically developing a technical framework of high-dimensional asymptotic analysis to derive the exact performance curve under various learning scenarios.