In the fall, I'll be joining Prime Intellect to scale their inference systems.
Previously, I was an intern on the Cloudflare Workers AI team, where I wrote some fast kernels and built secure systems. I'm also working on Tokenspeed, a speed-of-light LLM inference engine, with mentorship from Mingxing Zhang and Tsinghua MADSys group.
I spend my free time contributing to SGLang's Diffusion engine (I think diffusion is very cool!) while also learning about MLsys, distributed systems, and inference optimization."Once a word leaves your mouth, even four horses cannot chase it back." - 邓析
今年秋天，我将加入 Prime Intellect ，负责扩展他们的推理系统。
此前，我曾在 Cloudflare 的 Workers AI 团队实习，编写了一些高性能 kernel 并构建了安全系统。此外，我也在参与 Tokenspeed， 一个 speed-of-light LLM 推理引擎，并在 Mingxing Zhang 和 清华 MADSys 课题组的指导下工作。
我在空闲时间为 SGLang 的 Diffusion engine 做贡献（我觉得扩散模型非常酷！），同时学习机器学习系统、分布式系统和推理优化。一言既出，驷马难追
