Alham Aji
Visiting Research Scientist at MBZUAI (Mohamed bin Zayed University of Artificial Intelligence)
Abu Dhabi, Abu Dhabi Emirate, United Arab Emirates
Summary
🤩
Rockstar🎓
Top SchoolAlham Aji is a visiting research scientist and assistant professor with a decade of experience building efficient, multilingual NLP systems, specializing in low-resource and low-compute settings. He earned a PhD from the University of Edinburgh and has blended academic rigor with industry impact through roles at Google, Amazon, and Apple, plus substantial open-source contributions to Marian NMT—implementing parameter sharding, sparse tensor ops, and memory-saving techniques for multi-GPU translation. His research trajectory spans fast, compressed Transformer models and privacy-conscious, browser-based translation to practical deployments like Siri Malay and Alexa features. Based in Abu Dhabi, he co-organizes a community drive to advance Indonesian NLP, combining dataset and benchmark creation with hands-on mentoring of junior researchers. He is notable for translating cutting-edge ML research into production-aware optimizations that materially reduce training and inference cost.
10 years of coding experience
7 years of employment as a software developer
Doctor of Philosophy - PhD Natural Langauge Processing; Machine Translation, Doctor of Philosophy - PhD Natural Langauge Processing; Machine Translation at The University of EdinburghHigh School Science, High School Science at SMA N 1 DepokBachelor's Degree Computer Science, Bachelor's Degree Computer Science at University of Indonesia
Github Skills (12)
cuda10
sparse-matrix10
c-language10
sparse-data10
cprogramming-language10
gpu10
neural-machine-translation10
back-end-development10
multi-gpu10
multiple-gpu10
optimization10
machine-learning9
Programming languages (14)
Github contributions (5)
marian-nmt/marian
Feb 2017 - Nov 2020
Fast Neural Machine Translation in C++
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:62 commits, 3 pushes, 1 branch in 3 years 9 months
Contributions summary:Alham focused on enhancing the `marian` repository with parameter sharding techniques, aimed at improving performance on multi-GPU environments. Their contributions included the implementation of parameter sharding, the introduction of sparse tensor operations and its related operations, and optimizing data movement during the forward and backward passes. Further work was done on gradient dropping for reducing the memory footprint. Finally, the user also added support for moving average and exponential smoothing, providing a means to smooth out the model's training.
machine-translationtranslationneural-machine-translationcudagpu
afaji/afaji.github.io
Nov 2016 - Dec 2020
Contributions:1 PR, 56 pushes, 1 branch in 4 years 1 month
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.