Hangyu Zhang and Zaifu Zhan awarded doctoral dissertation fellowships
Congratulations to ECE’s Hangyu Zhang and Zaifu Zhan for winning the Graduate School’s 2025-2026 doctoral dissertation fellowship (DDF) award. The fellowship gives the University's most accomplished doctoral candidates an opportunity to devote full-time effort to an outstanding research project by providing time to finalize and write a dissertation during the fellowship year.
Hangyu Zhang
Hangyu Zhang is conducting his doctoral research under the guidance of Professor Sachin Sapatnekar (Distinguished McKnight University Professor, Robert and Marjorie Henle Chair). He is working on developing high performance placement methods for chip design. Current technologies from communications to automotive electronics to artificial intelligence (AI) rely increasingly on complex integrated circuits (ICs). Key to IC design is physical placement, which determines the spatial location of building blocks on a chip. Placement decisions can profoundly impact factors such as wirelength, timing, power consumption, thermal behavior, and manufacturability, which in turn can determine the cost and reliability of modern electronic products. For advanced ICs with millions to billions of placeable objects, placement must be performed using electronic design automation (EDA) algorithms, which rely on large-scale optimization algorithms. However, as designs grow more complex, traditional placement algorithms face challenges such as high computational cost, reliance on heuristic approximations, and limited scalability. Zhang’s research applies the power of algorithmic techniques and machine learning to improve the runtime of placement, while preserving the quality of the solution. Through insights gained from the integration of physics-based analytical placement with data-driven machine learning models and hardware acceleration, Zhang’s results have shown significant improvements over the state of the art.
Impact of Zhang’s work:
Improvements in physical design and EDA algorithms can indirectly affect billions of people by enabling faster, more energy-efficient, and more affordable computing systems. Better optimization methods can reduce power consumption, improve chip performance, and accelerate the development of emerging technologies such as AI, edge computing, autonomous systems, and advanced medical devices.
Zaifu Zhan
Zaifu Zhan is conducting his doctoral research under the supervision of Professor Rui Zhang (Chief, Division of Computational Health Sciences, University of Minnesota Medical School). He is working on developing large language models (LLMs) methods for efficient and accurate natural language processing (NLP) in biomedical settings. Biomedical knowledge often exists in text, in forms such as clinical notes, scientific articles, trial descriptions, drug labels, guidelines, and other similar forms. These are bodies of knowledge that are critical for research as well as for patient care. But its unstructured and terminology-heavy nature, complicated by its context-driven phrasing, frequent use of abbreviations, and local shorthand make it difficult to operationalize at scale for information extraction and summarization. Zhan’s research seeks to develop efficient and trustworthy LLM methods for biomedical NLP. The goal is to make high-performing models practical under constraints such as limited compute, privacy requirements, and the need for output that is grounded in evidence. Zhan’s research could support accurate decision making in healthcare settings (hospitals and/or research) and offer NLP systems that are not only accurate but also cost-effective, traceable, and easier to maintain.
Impact of Zhan’s work:
Zhan’s research has two key areas of impact. Firstly, it advances clinical and biomedical decision support by enabling faster, more reliable extraction, summarization, and question answering from clinical notes and biomedical literature. Secondly, it lowers the barrier to real-world deployment of biomedical LLMs through efficiency techniques that reduce compute, memory, and latency, making on-premise and privacy-preserving use more feasible for hospitals and research institutions.