일시: 2026. 3. 25. (수요일), 오후 4시

장소: 제4공학관 D507호

Speaker: Jeehoon Kang(강지훈), Ph.D. / Chief Research Officer (CRO) at FuriosaAI

Title: Unlock Furiosa RNGD's Full Potential with Kernel Programming

Abstract:

Furiosa RNGD delivers the predictable, high-performance inference necessary for today’s massive AI workloads. This architecture overcomes legacy GPU bottlenecks by utilizing a programmable dataflow for granular control over matrix and tensor operations, alongside 256MB of on-chip SRAM and 48GB of HBM for efficient data movement.
Furiosa Kernel Programming Model transforms RNGD into a foundational platform by granting developers direct, uncompromised access to the hardware. By bypassing generic compiler overhead through manual control over instruction scheduling and memory management, engineers can craft bespoke, high-efficiency kernels. Through this approach, we are building a vibrant ecosystem dedicated to pushing the boundaries of AI infrastructure.

Bio:

Dr. Jeehoon Kang is the Chief Research Officer (CRO) at FuriosaAI. Before joining the industry, he served as an Associate Professor of Computer Science at KAIST after studying under Professor Chung-kil Hur at Seoul National University. A specialist in concurrency, compilers, and software verification, he has published over 30 papers in premier venues such as SOSP, MICRO, ASPLOS, and PLDI.
At FuriosaAI, Dr. Kang leads the development of kernel programming models to maximize the performance of the RNGD accelerator, alongside building the application software stack for reinforcement learning and agentic AI. He is a key driver of the AI accelerator ecosystem, bridging the gap between cutting-edge academic research and industrial-scale deployment.