I'm a systems engineer working at the intersection of operating systems, hardware architecture, and AI infrastructure. My background is in Linux kernel development, with extensive work in storage systems, memory management, and filesystem infrastructure.
At Samsung Semiconductor I focus on the systems challenges of large-scale workloads, exploring how emerging compute and memory architectures shape AI inference performance. My recent work studies the limits of large language model inference, particularly KV-cache bandwidth constraints and kernel-level optimizations for long-context decoding.
Over two decades in systems engineering, I've contributed to core Linux subsystems and built open-source infrastructure used by kernel developers worldwide. I'm interested in how low-level system design influences the next generation of AI computing platforms and believe the best solutions emerge through open collaboration.
KV cache quantization, bandwidth scaling, and memory placement research for large language model inference.
Interactive visualization of KV cache bandwidth scaling across GPU architectures and transformer models.
NVMe QLC enablement through XFS large block size support in the Linux kernel.
Linux kernel development and testing automation framework.
Board Member of the Open Source Hardware Association.
Backport automation for Linux kernel drivers.
Central Regulatory Domain Agent for wireless regulatory compliance.
My work historically focused on Linux kernel development and systems infrastructure. Recently my research and focus has shifted toward the systems side of large-scale AI inference, particularly memory bandwidth constraints, KV cache architectures, and inference efficiency.
Inria — Coccinelle backporting research
Rutgers University