I’m currently an A.B. Duke Scholar at Duke University advised by Cynthia Rudin. Through my research, I hope to enable learning with less labels. This makes me particularly interested in representation learning, generative modeling, and self-supervised methods as well as their intersections.
In Fall 2020, I will be starting a PhD in Computer Science at Columbia University, exploring these topics. My doctoral work is supported by a Columbia Presidential Fellowship and an NSF Graduate Research Fellowship.
In my undergraduate work, I conducted research on applying generative models to other machine learning tasks (such as image super-resolution and reinforcement learning), as well as on fundamental techniques developing such models (especially normalizing flows). In industry, I have spent time working on data science/machine learning for neurotech and for cold storage logistics.
Feel free to reach out!
PhD in Computer Science
BS in Mathematics & Computer Science, 2020
Current super-resolution methods optimize on pixel-wise average correctness measures which lead to blurring. We present an alternative problem formulation that focuses on creating perceptually realistic images that downscale correctly. Our algorithm, PULSE, solves this problem by doing a self-supervised search of the outputs of a generative model for images that downscale correctly, leveraging some properties of high-dimensional Gaussians; this yields far better perceptual quality than previous methods.