I am a final-year Ph.D. student at NYU advised by Leon Bottou and Yann LeCun. I have broad research interests in machine learning, representation learning and out-of-distribution generalization. Recently, I am interested in the learning principles to build AI for the open-world.  AI for the open-world aims at solving any task that a human could possibly undertake. That is, a wide range of tasks with fewer examples and less priori knowledge from human designers. I explores three such learning principles, inluding rich features, disentangled representation, and inference-time learning  to support each of the three requirements above.