Harry H. Zhang
I am a graduate student in the SPARK Lab of MIT LIDS. I am extremely fortunate to be advised by Prof. Luca Carlone.
Prior to MIT, I was a MS-Research student in the CMU Robotics Institute studying Artificial Intelligence and Robotics, advised by Prof. David Held. I also worked at Amazon as an Applied Scientist II.
Prior to CMU, I earned my B.S. (2017–2021) with Honors from UC Berkeley with a major in EECS and a minor in Mechanical Engineering. During my time at Berkeley, I did research under Prof. Ken Goldberg and Dr. Jeffrey Ichnowski in AUTOLab. I maintain and curate a popular deep reinforcement learning tutorial on my GitHub.
Research Interests
My background in robot learning naturally leads me to quantitative finance, where I develop systematic trading strategies and study how predictive signals can be translated into robust, monetizable alpha. My work in quantitative finance is a natural extension of my research in robot learning and trustworthy deep learning: both require extracting signal from noisy, high-dimensional data, reasoning carefully about uncertainty and distribution shift, and building learning systems that remain reliable outside the regimes in which they were developed.
In many ways, quantitative research and robot learning pose remarkably similar problems. A robot must learn from imperfect observations of a complex environment, form beliefs about latent states, quantify what it does not know, and make sequential decisions whose consequences are only observed afterward. A trading system faces much the same challenge: markets are partially observed and nonstationary, observations are noisy, actions affect realized outcomes, and the relationship between historical data and the future is inherently uncertain. In both settings, strong in-sample prediction is relatively easy; the harder problem is determining which learned relationships will survive distribution shift and remain useful when deployed.
Research Highlights
H2OFlow
ICLR 2026
Grounding 3D human-object affordances using generative models and dense diffusion flows.
CUPS
ICML 2025
Improving human pose-shape estimation with conformalized deep uncertainty.
TAX-Pose
CoRL 2022
Task-specific cross-pose estimation for generalizable robot manipulation.
FlowBot3D
RSS 2022 — Best Paper Finalist
Learning 3D articulation flow to manipulate novel articulated objects.