Data-efficient learning for real-world vision: few-shot, semi-supervised, and robust generalization
I work on few-shot image classification, semi-supervised learning, graph-based representation learning, vision transformers, and medical imaging applications—building methods that generalize under limited labels and domain shift.
Core Focus
8+ years in CV research and 10+ years building production systems. I design learning frameworks that remain stable when labels are scarce.
Research
Topics I actively work on and the type of problems I like to solve.
Label-efficient recognition
Few-shot & semi-supervised learning pipelines that reduce label cost and remain reliable under distribution shift.
Transformers + Graph reasoning
Token refinement, uncertainty-aware selection, and graph propagation to exploit intra-/inter-class structure.
Medical imaging applications
Data-scarce clinical settings: robust models, explainability, and safe evaluation protocols.
Practical engineering
Strong software background for building reproducible experiments and deployable research prototypes.
Publications
Teaching
Selected teaching roles.
Service
Editorial & peer-review service.
IEEE Transactions on Multimedia; Applied Soft Computing; Applied Intelligence; Digital Signal Processing; IEEE Access; Scientific Reports; IJCNN (2025–2026).
Reproducible experimentation, careful evaluation, and clear scientific writing—aligned with high-standard venues.
Contact
Let’s collaborate or discuss opportunities.
ORCID: 0000-0003-3651-4954