Ali Hassani
I am a computer science PhD student at Georgia Tech, and am advised by Prof. Humphrey Shi. I received my master's degree from University of Oregon under Prof. Shi, and my undergraduate from University of Kerman, where I did research under Prof. Abbas Salemi. I'm interested in a wide range of topics concerning Deep Learning, Computer Vision, and High Performance Computing. My research started out with classical unsupervised ML methods during my undergraduate, and deep learning and computer vision in PhD, and shifted into machine learning systems, system co-design, and high performance computing. I've been fortunate to have interned at HippoML (under Bing Xu) and NVIDIA (mentored by Haicheng Wu). My goal is to contribute to help scale up current and future AI and general-purpose computing systems. My longest running research project is Neighborhood Attention, a multi-axis sliding window attention pattern that allows fine-grained control over attention span, drawing a spectrum of possible attention patterns between linear projection and self attention. Most of my research has involved designing deep learning architectures based on Neighborhood Attention, scaling them up, and most importantly, accelerating it for general-purpose accelerators, namely NVIDIA GPUs. Check out our open-source project NATTEN for more details. I've also contributed to open source projects such as CUTLASS.
GitHub / Google Scholar / Twitter / LinkedIn
