My research interests lie in efficient and resource-aware machine learning: continual learning, knowledge distillation and model compression, and compact representation learning.
I am a research assistant at the VinUni–Illinois Smart Health Center. In 2026 I was a visiting researcher with Prof. S. Kevin Zhou at USTC (MIRACLE), where the work below began. I finished a B.Sc. in Data Science at the University of Information Technology, VNU-HCM, as co-valedictorian.
Author in bold. Manuscripts under review are marked as such.
Efficient & continual learning
Nguyen Ngoc Yen Nhi, M.-T. Tran, Y. Li, H.-H. Pham, S. K. Zhou
Under review, ICLR 2027
note paper (under review)
D. H. Nguyen, Nguyen Ngoc Yen Nhi, H.-H. Pham
GlobalSouthML Workshop, ICML 2026
D. H. Nguyen, Nguyen Ngoc Yen Nhi, H.-H. Pham
Workshop on Structured Data for Health, ICML 2026
Multimodal, retrieval & medical AI
Nguyen Ngoc Yen Nhi, T. Nguyen, K. H. C. Tuan, H.-H. Pham
ODIN Challenges, MICCAI 2026
G. S. Nguyen, Nguyen Ngoc Yen Nhi, W. Buntine, D. D. Le
BlackboxNLP Special Track, EMNLP 2026
A primer: how a frozen backbone and a single closed-form equation let a model learn new classes with no replay and no gradients. Builds from the problem up through ACIL and RanPAC to the memory question my work asks.
Read the primer Paper note · under reviewThe question behind my ICLR 2027 submission: how should an analytic continual learner split a fixed memory budget between width and interaction fidelity? Notes on the problem — the method and results stay out while the paper is under review.
Read the note