Hayate Okuhara

About

Hayate Okuhara is a leading researcher in energy-efficient AI-IoT System-on-Chip (SoC) design, with a focus on heterogeneous computing architectures for extreme edge devices. His work addresses the critical challenge of running diverse, compute-intensive Deep Neural Network (DNN) workloads—from augmented reality to nano-robotics—within a power envelope of just tens of milliwatts. Okuhara’s major contribution is the development of precision-scalable DNN accelerators that dynamically adjust bit-widths from 2 to 8 bits, enabling optimal trade-offs between accuracy and energy efficiency. His flagship design, the AI-IoT SoC featuring 16 RISC-V cores, achieves a remarkable 12.4 TOPS/W at 136 GOPS, as detailed in his most-cited paper (34 citations). A standout innovation is his adaptive body biasing technique, which boosts performance by 30% under varying operating conditions. Okuhara’s work, exemplified by the Marsellus SoC, demonstrates how heterogeneous RISC-V architectures can deliver both high throughput and ultra-low power consumption, setting a new standard for intelligent edge computing.

Research Focus

Key Achievements

Top Papers

1. 1
   22.1 A 12.4TOPS/W @ 136GOPS AI-IoT System-on-Chip with 16 RISC-V, 2-to-8b Precision-Scalable DNN Acceleration and 30%-Boost Adaptive Body Biasing

   34 citations · 2023
2. 2
   Marsellus: A Heterogeneous RISC-V AI-IoT End-Node SoC with 2-to-8b DNN Acceleration and 30%-Boost Adaptive Body Biasing

   2 citations · 2023

Key Collaborators

Contact & Links