Osama Yousuf

Graduate Research Assistant osamayousuf@gwu.edu

I am a PhD candidate in Computer Engineering at The George Washington University, specializing in Machine Learning & Intelligent Systems. My research focuses on deep learning in alternative computing paradigms (neuromorphic & in-memory computing), aiming to make deep neural network training and inference more resource-efficient through algorithmic and architectural innovations.

I bring a unique mix of software and hardware engineering expertise. I can navigate deep learning frameworks 🧠💡, design AI/ML models with hardware-software co-design constraints 🛠️⚡, and develop simulation libraries for AI accelerators and intelligent systems 🌐🤖.

I graduate in Spring 2025 and am actively seeking roles in AI/ML research and engineering. I am excited to join a team that pushes the boundaries of AI and challenges me to grow!

The best way to reach me is via e-mail.

news

Feb 01, 2025	algorithmic work on a fault-tolerant inference scheme for noisy neural networks published in Nature Communications: Neuromorphic Engineering
Jan 30, 2025	successfully defended my PhD dissertation titled “Memristive Neural Networks: Modeling, Prototyping, and Hardware-Software Co-Design”!
Jan 17, 2025	filed patent on a general fault-tolerance scheme for neural network accelerators based on noisy hardware
Dec 19, 2024	delivered a seminar titled “Accelerating LLMs with Magnetic Crosspoint Arrays” at Western Digital, San Jose, CA, concluding my research internship with the Advanced Memory Technology group
Oct 19, 2024	presented work on an ensembling technique for added fault-tolerance in noisy neural networks at the Innovation Bazaar, Western Digital, Milpitas, CA

selected publications

Nature

Layer ensemble averaging for fault tolerance in memristive neural networks

Osama Yousuf, Brian D Hoskins, Karthick Ramu, Mitchell Fream, and 7 more authors

Nature Communications, 2025

DOI Bib

@article{yousuf2025layer,
  title = {Layer ensemble averaging for fault tolerance in memristive neural networks},
  author = {Yousuf, Osama and Hoskins, Brian D and Ramu, Karthick and Fream, Mitchell and Borders, William A and Madhavan, Advait and Daniels, Matthew W and Dienstfrey, Andrew and McClelland, Jabez J and Lueker-Boden, Martin and others},
  journal = {Nature Communications},
  volume = {16},
  number = {1},
  pages = {1250},
  year = {2025},
  publisher = {Nature Publishing Group UK London},
  doi = {10.1038/s41467-025-56319-6},
  url = {https://www.nature.com/articles/s41467-025-56319-6},
  dimensions = {true},
}

NANOARCH

Neural Network Modeling Bias for Hafnia-based FeFETs

Osama Yousuf, Imtiaz Hossen, Andreu Glasmann, Sina Najmaei, and 1 more author

In Proceedings of the 18th ACM International Symposium on Nanoscale Architectures, 2023

DOI Bib

@inproceedings{yousuf2023neural,
  title = {Neural Network Modeling Bias for Hafnia-based FeFETs},
  author = {Yousuf, Osama and Hossen, Imtiaz and Glasmann, Andreu and Najmaei, Sina and Adam, Gina C},
  booktitle = {Proceedings of the 18th ACM International Symposium on Nanoscale Architectures},
  pages = {1--5},
  year = {2023},
  doi = {10.1145/3611315.3633253},
  url = {https://dl.acm.org/doi/abs/10.1145/3611315.3633253},
  dimensions = {true},
}

Frontiers

Gradient decomposition methods for training neural networks with non-ideal synaptic devices

Junyun Zhao, Siyuan Huang, Osama Yousuf, Yutong Gao, and 2 more authors

Frontiers in Neuroscience, 2021

DOI Bib

@article{zhao2021gradient,
  title = {Gradient decomposition methods for training neural networks with non-ideal synaptic devices},
  author = {Zhao, Junyun and Huang, Siyuan and Yousuf, Osama and Gao, Yutong and Hoskins, Brian D and Adam, Gina C},
  journal = {Frontiers in Neuroscience},
  volume = {15},
  pages = {749811},
  year = {2021},
  publisher = {Frontiers Media SA},
  doi = {10.3389/fnins.2021.749811},
  url = {https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2021.749811/full},
  dimensions = {true},
}

ICONS

A system for validating resistive neural network prototypes

Brian Hoskins, Wen Ma, Mitchell Fream, Osama Yousuf, and 7 more authors

In International Conference on Neuromorphic Systems 2021, 2021

DOI Bib

@inproceedings{hoskins2021system,
  title = {A system for validating resistive neural network prototypes},
  author = {Hoskins, Brian and Ma, Wen and Fream, Mitchell and Yousuf, Osama and Daniels, Mathew and Goodwill, Jonathan and Madhavan, Advait and Tung, Hoang and Branstad, Mark and Liu, Muqing and others},
  booktitle = {International Conference on Neuromorphic Systems 2021},
  pages = {1--5},
  year = {2021},
  doi = {10.1145/3477145.3477260},
  url = {https://dl.acm.org/doi/abs/10.1145/3477145.3477260},
  dimensions = {true},
}