Ali Hassani
Research Scientist at NVIDIA
San Francisco Bay Area United States
Summary
🤩
Rockstar🎓
Top SchoolAli Hassani is a Research Scientist at NVIDIA with five years of experience building high-performance AI architectures for generative and physical AI. He couples deep academic training (PhD work at Georgia Tech and prior research at University of Oregon) with hands-on systems expertise, having implemented CUDA kernels to accelerate Neighborhood Attention Transformers used in CVPR-class research. Based in the San Francisco Bay Area, Ali has progressed from internships and teaching roles to production-focused research, bridging algorithmic innovation and low-level performance engineering. His background uniquely blends GPU optimization, transformer attention mechanics, and practical deployment experience, making him adept at turning cutting-edge models into efficient, scalable implementations.
5 years of coding experience
4 years of employment as a software developer
Master of Science - MS Computer Science, Master of Science - MS Computer Science at University of OregonDoctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at Georgia Institute of TechnologyBachelor's degree Computer Science, Bachelor's degree Computer Science at University of Kerman
English, Persian
Github Skills (5)
cuda10
pytorch10
machine-learning10
deep-learning10
computer-vision9
Programming languages (8)
Github contributions (5)
SHI-Labs/Neighborhood-Attention-Transformer
Apr 2022 - Jan 2023
Neighborhood Attention Transformer, arxiv 2022 / CVPR 2023. Dilated Neighborhood Attention Transformer, arxiv 2022
Role in this project:
ML Engineer
ML EngineerContributions:70 commits, 30 PRs, 19 pushes in 9 months
Contributions summary:Ali's initial commit introduces CUDA kernels for the Neighborhood Attention Transformer, showcasing expertise in optimizing attention mechanisms for image processing. The changes involve the creation of CUDA kernels for forward and backward passes, along with the integration of relative position biases (RPB), indicating focus on efficient implementation for deep learning models. The commit includes detailed code, demonstrating proficiency in CUDA programming for accelerating matrix operations and implementing model components specific to the described transformer architecture.
pytorchattention-mechanismneighborhoodtransformerself-attention
alihassanijr/cutlass
Dec 2022 - Oct 2024
CUDA Templates for Linear Algebra Subroutines
Contributions:66 pushes, 20 branches in 1 year 11 months
cudalinear-algebrandarraygpulinear
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.