Before joining the residency, I completed my Bachelor’s and Master’s degrees in computer science at MIT. I got my first taste of research during a summer internship at an aerospace company where I worked on anomaly detection algorithms for time-series data. Since then, I’ve pursued projects in a variety of disciplines, including numerical optimization, computational connectomics, and computer vision. Many of my projects involved machine learning, and I frequently used neural nets as black-box feature extractors for esoteric problems like “occluded pose estimation.” However, I never developed a firm intuition for why they work so well, so I applied for the AI Residency program in hopes to broaden my understanding. The experience so far has been incredible. Residents are given the freedom to pursue their own research directions with guidance from experts in the field. There are countless avenues for learning and growth -- reading groups, invited talks, internal courses, to name a few. I’m currently studying the ways in which data influences the optimization landscape of deep nets and how we can leverage these insights to train neural nets faster. My goal for the year is to get us one step closer to a “science of deep learning,” or at least learn something new along the way!