Research

My research asks a practical question: how can institutions learn together without giving up control of sensitive data?

Federated learning

I study client selection, asynchronous optimization, non-IID data, fairness, communication-aware learning, and adversarial robustness. The goal is federated systems that remain effective—and secure—under the constraints of real networks and organizations.

Graph learning

Financial activity is relational: accounts, devices, merchants, transactions, and institutions form connected systems. My work explores graph representation learning, graph contrastive learning, and personalized federated graph transformers.

AI for financial security

I develop AI methods for trustworthy financial applications, including credit-data synthesis, default prediction, privacy-preserving collaboration, risk intelligence, and retrieval-augmented language models where factual reliability matters.

Learning security

My recent work investigates multi-target backdoor attacks and the robustness of mainstream defenses in federated learning, connecting privacy-preserving collaboration with the broader security requirements of deployed AI.

Edge intelligence

My earlier and continuing work studies computation offloading and over-the-air learning, connecting model design to the communication and computing infrastructure on which distributed intelligence runs.

Signal processing

Signal processing provides a foundation for my work on distributed sensing, hybrid beamforming, data censoring, over-the-air computation, and communication-efficient federated learning.