Hengguan Huang

Learn explicit latent mechanisms—events, graphs, and dynamics—that explain how observations are generated.

I develop Bayesian deep models that recover structured latent mechanisms from incomplete observations, combining probabilistic graphs, continuous-time dynamics, and neural representations.

Build probabilistic and agentic systems that seek evidence, compare hypotheses, and revise beliefs under uncertainty.

I combine Bayesian inference, active evidence acquisition, and language-model reasoning to build scientific agents that know what to observe next, when to abstain, and how to update explanations.

Develop calibrated, fair, robust, and safe AI for high-stakes health systems under population and distribution shift.

I study uncertainty calibration, test-time adaptation, fairness, and generative modeling so clinical and public-health AI remains dependable as populations, modalities, and evidence change.