Dongxia Wu
Assistant Professor of Statistics and Data Science
Research Interests
Professor Wu’s teaching and research interests span the intersection of trustworthy AI and AI for science, with a focus on uncertainty-aware scientific modeling, generalization under distribution shift, and biologically and physically grounded modeling. He is also interested in sample-efficient scientific discovery and reliable decision-making and control. His goal is to develop machine learning models and frameworks that are robust, generalizable, well-calibrated, and scientifically meaningful across a wide range of scientific applications.
Email
Prior to joining MBZUAI, Professor Wu was a postdoctoral researcher at Stanford University, where he worked on deep generative modeling, uncertainty quantification, and reinforcement learning to develop trustworthy frameworks for single-cell modeling, with the goal of enabling more biologically plausible, uncertainty-calibrated, and reliable models for scientific discovery. He received his Ph.D. in Computer Science from University of California San Diego in 2025. His previous research focused on probabilistic machine learning methods for large-scale, high-dimensional, and structured data, including time series, spatiotemporal, and language data, with applications in public health, climate science, and drug discovery. Professor Wu also works on uncertainty-guided control and optimization.
- Postdoctoral Fellow, StanfordUniversity, USA
- Ph.D. in Computer Science, University of California San Diego, USA
- B.Sc. in Applied Mathematics, University of Wisconsin–Madison, USA
- CSE Dissertation Award, Runner-up, UC San Diego, 2026
- Keynote Speaker, USC Symposium on Frontiers of ML/AI, 2025
- HDSI Ph.D. Fellowship, University of California San Diego, 2021-2024
- Developed DeepGLEAM for COVID-19 incident death forecasting, achieving the highest coverage ranking in the CDC Forecasting Hub, 2021
- Eckmann P, Wu D, Heinzelmann G, Gilson MK, and Yu R.: “MF-LAL: Drug compound generation using multi-fidelity latent space active learning”, ICML, 2025.
- Niu R, Wu D, Kim K, Ma YA, Watson-Parris D, and Yu R.: “Multi-fidelity residual neural processes for scalable surrogate modeling”, ICML, 2024.
- Wu D, Niu R, Chinazzi M, Ma YA, and Yu R.: “Disentangled multi-fidelity deep Bayesian active learning”, ICML, 2023.
- Wu D, Niu R, Chinazzi M, Vespignani A, Ma YA, and Yu R.: “Deep Bayesian active learning for accelerating stochastic simulation”, KDD, 2023.
- Wu D, Chinazzi M, Vespignani A, Ma YA, and Yu R.: “Multi-fidelity hierarchical neural processes”, KDD, 2022.
- Wu D, Gao L, Xiong X, Chinazzi M, Vespignani A, Ma YA, and Yu R.: “Quantifying uncertainty in deep spatiotemporal forecasting”, KDD, 2021.