I'm a researcher at Hudson River Trading, where I develop state of the art deep learning systems for trading on HRT AI Labs (HAIL).
I completed my SB and MEng in Computer Science at MIT in 2023, advised by Devavrat Shah in the Laboratory for Information and Decision Systems. My graduate work centered on time series analysis, matrix completion, and statistical learning. Earlier, I collaborated with researchers at MIT and Google Brain on differentiable simulation for photovoltaic cells, with applications to inverse design and neural network surrogate models. I am broadly interested in the intersection of deep learning, modeling, and optimization.
I grew up in Hong Kong, and went to school on Argyle Street for over a decade.
SAMoSSA: Multivariate Singular Spectrum Analysis with Stochastic Autoregressive Noise
A two-stage time series method that applies multivariate singular spectrum analysis to extract non-stationary structure, then fits an autoregressive model to the residual, with finite-sample forecasting guarantees.
Exploiting Observation Bias to Improve Matrix Completion
A matrix completion framework that treats non-uniform observation patterns as signal rather than nuisance, recovering latent factors shared between the missingness mask and the underlying matrix.
∂PV: An End-to-End Differentiable Solar-Cell Simulator
A fully differentiable photovoltaic simulator built on drift-diffusion physics that enables gradient-based inverse design and neural network surrogate modeling of solar cells.
AI Researcher
Hudson River Trading
Member of HRT AI Labs (HAIL), building deep learning systems for trading.
Massachusetts Institute of Technology
Advisor: Devavrat Shah · LIDS · MEng thesis: SAMoSSA (see Publications)