Hi, nice to meet you! I am Zhijing Li, a software engineer in Pytorch accelerator team at Meta. Our recent focus is AITemplate, a python framework that transforms AI models into NVIDIA and AMD GPU with best performance.

I used to be a Phd candidate at Cornell University, but graduated with a Master degree. My advisor was Prof. Adrian Sampson in Capra research group. While I graduated early, it’s because my interest can also be fufilled in industry and Adrian is one of the best advisors and most respectable people I ever knew.

During my graduate school, my research focus was on programming languages and compilers for designing AI accelerators, where I designed EQueue dialect in MLIR and Magnolia as simulation language for AI accelerators. I also worked on image compression for deep neural networks applications, and contributed to Dahlia and FuTil.

Before coming to Cornell, I worked on approximate computing and stochastic computing with Prof. Weikang Qian at Shanghai Jiao Tong University.