Gateway South 433
1 Castle Point Terrace, Hoboken, NJ 07030
My name is Pengfei Hu. I’m a 3rd year Ph.D. student in Department of Computer Science at Stevens Institute of Technology, advised by Prof. Yue Ning. Before that, I obtained my master’s degree from the Viterbi School of Engineering, University of Southern California in 2023, where I worked with Prof. Sze-chuan Suen.
My current research concentrates on predictive healthcare on electronic health records (EHR), with a specific focus on augmented (e.g. knowledge-guided and retrieval augmented) language models and their robustness (e.g. generalization and transparent adaptation) under domain shifts. I have also worked on graph neural networks, LLM Agents, and Tool-Integrated Reasoning designs.
Feel free to drop me an email (phu9 at stevens dot edu) if you have any questions about my research, or want to discuss about potential collaborations.
Educations
Industrial Experience
- Oak Ridge National Laboratory (May 2025- Aug. 2025)
- Graduate Research Intern, Computational Sciences and Engineering Division
- Topic: Reservoir Inflow Forecasting. [Two Workshop and One Journal papers]
- Mentor: Fan Ming
News
| Jun 16, 2026 | I will be joining the AWS AI Lab as an Applied Scientist Intern this fall! |
|---|---|
| May 04, 2026 | Our paper II-KEA won the Best Doctoral Research Poster Award at the iCNS @ Stevens AI Engineering and Science Symposium. |
| May 01, 2026 | Our paper Exploring Accurate and Transparent Domain Adaptation in Predictive Healthcare via Concept-Grounded Orthogonal Inference is accepted to ICML 2026. See you in Seoul! |
| Apr 17, 2026 | I passed my Thesis Proposal defense. |
| Feb 27, 2026 | Our paper AGFormer: Adaptive Spatiotemporal graph informed transformer for multi-reservoir inflow forecasting is accepted to Environmental Modelling & Software (IF: 4.6). |
Selected Publications
- ICML 2026
Exploring Accurate and Transparent Domain Adaptation in Predictive Healthcare via Concept-Grounded Orthogonal InferenceAccepted by the 43rd International Conference on Machine Learning (ICML), 2026 - arXiv
UdonCare: Hierarchy Pruning for Unseen Domain Discovery in Predictive HealthcarearXiv preprint arXiv:2506.06977, 2025