School of Electrical Engineering and Computer Science (EECS) researchers Pratyush Dhingra, Pramit Kumar Pal, Jana Doppa, and Partha Pratim Pande recently received a best paper award at the IEEE/ACM Embedded Systems Week (ESWEEK) event in Barcelona, Spain.
The WSU team received the award for their paper “ThAME: 3D Memory-Enabled Heterogeneous Accelerator for LLM Mixture of Experts.” Dhingra, a Ph.D. student in the School of EECS, was the lead author.
“It’s incredibly rewarding. ESWEEK is one of the premier venues for embedded and hardware systems research, and receiving the Best Paper Award means a lot to me,” Dhingra said. “As I come to the end of my Ph.D., this recognition truly feels like the summation of years of work and expertise developed across hardware design, AI, and system optimization.”
Dhingra said AI models such as ChatGPT and Claude have adopted “Mixture of Experts” (MoE) designs. Running these models can be slow and power-hungry because standard graphics processing units spend much of their time moving large amounts of data between memory and processors. “Our paper proposes a new kind of hardware architecture that stacks memory and computing closely together, paired with a specialized communication backbone, allowing these models to respond much faster while being significantly more energy efficient,” said Dhingra.
He credited his advisor, Partha Pande, for his guidance and encouragement throughout his doctoral studies and thanked his lab mates and the School of Electrical Engineering and Computer Science for providing a supportive research environment.
