Caiwen Ding wins NSF CAREER Award
![Caiwen Ding headshot](/sites/cse.umn.edu/files/styles/webp_scaled/public/Image_20240307231133.jpg.webp?itok=XFmn0lmo)
Department of Computer Science & Engineering’s (CS&E) Caiwen Ding has received a Faculty Early Career Development (CAREER) Award from the National Science Foundation (NSF) to support his project, “Algorithm-Hardware Co-design of Efficient Large Graph Machine Learning for Electronic Design Automation”. This prestigious award provides support to early-career faculty who have the potential to serve as academic role models in research and education and to lead advances in the mission of their department or organization. Ding will join CS&E this fall as an associate professor with a research focus on artificial intelligence and computing systems.
“My CAREER project focuses on the development of semiconductor technologies,” said Ding. “We are using machine learning as a tool to speed up the Electronic Design Automation (EDA) process for chip design. There are a number of limitations in the chip design process using classical analytical methods, especially for complex problems. Our methods aim to save time and computational resources for the multiple stages of the chip design process.”
The goal of Ding’s project is to address the efficiency and scalability of using graph neural networks (GNNs) for EDA through a series of algorithm-hardware co-design approaches. Processing large circuit graphs on mainstream computers is challenging, and often requires multi-GPU, which results in huge amounts of power consumption and high costs. Ding’s work will address the inefficiency and scalability of this technique to make processing large circuit graphs more accessible to the average researcher, and reduce the environmental cost of computation.
“We are studying graph node and edge features to represent the electronic circuits so that we can have good prediction accuracy,” said Ding. “Then we use approximation techniques to enable single GPU processing. We design GPU kernels tailored to the circuit graphs to speed up GNN training. Overall, we combine techniques ranging from algorithms to hardware to address the efficiency and scalability.”
In August 2022, the CHIPS and Science Act put more resources towards research and workforce training in chip design and manufacturing in the United States. From the teaching perspective, Ding is developing courses that combine machine learning with these semiconductor technologies. He hopes to collaborate with the Electrical and Computer Engineering (ECE) department in this endeavor. The goal is to give students the skills they need to be competitive in this growing job market.
“Under the umbrella of the strong CS&E and ECE programs at the University of Minnesota, I hope this award can foster exciting machine learning for EDA research directions,” said Ding. “I have already started the collaboration with several colleagues, and am looking forward to more.”
In addition to his collaborators at the U of M, Ding is working with Cunxi Yu from the University of Maryland, as well as EDA companies. He also credits his students and PhD advisor.
“I want to thank my students for their hard work and dedication to the research, and my PhD advisor, Yanzhi Wang, for the guidance, support, and mentorship throughout my doctoral studies.” said Ding.