Yangmin Liam Li

About Yangmin Liam Li

I’m a CMU graduate student passionate about building reliable and efficient systems at the intersection of AI and infrastructure.At NVIDIA, I work on the DriveWorks platform, developing custom memory-efficient tools for data processing in autonomous vehicle systems.Previously at ByteDance and AWS China, I gained hands-on experience in distributed systems and large-scale engineering.I’ve contributed to SGLang and Prism, a collaborative project with UC Berkeley that enables GPU-level memory sharing across 50+ LLMs using CUDA Virtual Memory.I also work on physics-grounded reasoning systems and hybrid LLM architectures, combining symbolic methods with reinforcement learning and scalable training.I enjoy working across both research and production — where ideas meet systems that scale.

Experience

1. Software Engineer

   NVIDIA

   Jan 2025 — Present · Santa Clara County, CA, US

   Contribute and optimize large-scale LLM inference systems at NVIDIA, focusing on high-performance serving, scheduling, and multi-GPU / multi-node deployment. Worked on SGLang and TensorRT-LLM related infrastructure to improve throughput, latency, and long-context serving efficiency on NVIDIA GPU platforms.

Education

• Tongji University

  Bachelor's degree, Computer Science

• Carnegie Mellon University

  Master of Science - MS, Information Technology | Information Security

  2023 — 2024