Sanghani Center Student Spotlight: Shuaicheng Zhang

At the 2025 ACM SIGKDD & Annual KDD Conference in August, Shuaicheng Zhang and collaborators from Massachusetts Institute of Technology and IBM garnered a Best Paper Award in the Benchmark and Dataset Track for "When Heterophily Meets Heterogeneity: Challenges and a New Large-Scale Graph Benchmark."  

“Our H2GB benchmark helps detect fraudulent transactions and money laundering by analyzing temporal patterns in financial transaction networks,” said Zhang. “We hope it provides new fertile ground for researchers still passionately pursuing graph learning, offering both the scale and complexity needed to tackle real-world challenges.”

A second paper, “MentorPDM: Learning Data-Driven Curriculum for Multi-Modal Predictive Maintenance,” on which Zang was first author, was also accepted in proceedings of the 2025 KDD conference. MentorPDM applies graph learning to industrial predictive maintenance, helping manufacturers predict equipment failures before they occur by analyzing multi-sensor time-series data.

Zhang, a Ph.D. student at the Sanghani Center advised by Dawei Zhou, said his research focus is driven by the challenge of working with complex relational data that mirrors real-world interconnected systems; the potential impact on critical applications in finance, healthcare, and industrial systems; and an interest in bridging theoretical innovations with practical applications with machine learning.

Among his other published papers are:

• "UNIFIEDGT: Towards a Universal Framework of Transformers in Large-Scale Graph Learning," proceedings of 2024 IEEE  International Conference on Big Data 

• "Personalized Federated Learning under Mixture of Distributions,"  proceedings of 2023 International Conference on Machine Learning (ICML)

• "TGEditor: Task-Guided Graph Editing for Augmenting Temporal Financial Transaction Networks," proceedings of 2023 ACM International Conference on AI in Finance (ICAIF)

Another paper, "HeroFilter: Adaptive Spectral Graph Filter for Varying Heterophilic Relations," has been accepted in proceedings at the upcoming 2025 NeurIPS conference in December.

Zhang earned a bachelor’s degree in computer science from Virginia Tech and a master’s degree in telecommunications from the University of Maryland, College Park. 

He said he was drawn back to Virginia Tech and the Sanghani Center when looking for a Ph.D. program because of their strong reputation in computer science and artificial intelligence; the university’s super computing resources; and a very positive undergraduate experience.

Projected to graduate in Spring 2026, Zhang is planning to pursue a career as an industrial research scientist.