Mostafa Elhoushi
Research Scientist at Cerebras Systems
Toronto, Ontario, Canada
Summary
🤩
Rockstar🎓
Top SchoolMostafa Elhoushi is a research scientist and machine learning engineer specializing in the intersection of deep learning, compilers, and systems, with over a decade of experience across AI infrastructure and hardware-aware optimization. He has driven production and research wins at Cerebras, Meta (FAIR), and Huawei—delivering runtime speedups, memory reductions, and novel sparsity and pruning methods adopted by the community and industry. His work spans building custom compiler languages and kernels for AI accelerators, ML-driven compiler-processor co-design, and practical model compression techniques, backed by patents and multiple ICML/CVPR publications. Comfortable moving between low-level C++/CUDA and high-level model research, he also mentors data curation that improved CLIP retrieval and led collaborations with top universities. Based in Toronto, he combines academic rigor from a PhD track with hands-on systems engineering, uniquely positioning him to close the gap between ML algorithms and hardware execution.
11 years of coding experience
13 years of employment as a software developer
MSc Computer and Systems Engineering, MSc Computer and Systems Engineering at Ain Shams UniversityPhD Electrical and Computer Engineering, PhD Electrical and Computer Engineering at Queen's University
English, Arabic
Stackoverflow
4,516reputation
2.8mreached
15answers
1question
cuda
top-1%
Github Skills (84)
transformers10
language-model10
python10
deep-learning10
gpu10
gemma10
quantization10
gpu-acceleration10
glm10
multimodal10
numpy10
audio10
neural-network10
memory-optimization10
transformer10
Programming languages (6)
Github contributions (5)
facebookresearch/LayerSkip
Oct 2024 - May 2025
Code for "LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding", ACL 2024
Contributions:21 reviews, 7 PRs, 27 pushes in 6 months
inferencellmoptimizationtransformersearly-exit
mostafaelhoushi/tensor-decompositions
Jun 2019 - May 2020
PyTorch implementation of [1412.6553] and [1511.06530] tensor decomposition methods for convolutional layers.
Contributions:139 commits, 62 pushes in 11 months
pytorchtensor
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.