Ziyang Shen

About

Ziyang Shen is a rising star in the field of energy-efficient neuromorphic computing and embodied artificial intelligence. His research focuses on bridging the gap between advanced vision transformer architectures and low-power hardware accelerators, particularly for real-time autonomous systems. In his most-cited work, "A 28nm Spiking Vision Transformer Accelerator with Dual-Path Sparse Compute Core and EMA-free Self-Attention Engine for Embodied Intelligence," Shen introduces a novel hardware design that integrates spiking neural networks with transformer models to achieve dramatic energy savings without sacrificing performance. This accelerator, fabricated in 28nm CMOS, addresses critical bottlenecks in embodied AI—such as high power consumption and latency—by employing dual-path sparse computation and an exponential moving average-free self-attention engine. Though early in his career, Shen’s contributions are already recognized for their potential to enable next-generation autonomous robots and interactive agents that process visual data in real time. His work exemplifies the convergence of algorithm and hardware co-design, positioning him as a key innovator in the push toward truly intelligent, energy-sustainable embodied systems.

Research Focus

Key Achievements

Top Papers

1. 1
   A 28nm Spiking Vision Transformer Accelerator with Dual-Path Sparse Compute Core and EMA-free Self-Attention Engine for Embodied Intelligence

   2 citations · 2025

Key Collaborators

Contact & Links