Bilal Khan
Research Engineer at Isomorphic Labs
Canada
Summary
🤩
Rockstar🎓
Top SchoolBilal Khan is a research engineer with nine years of experience building and optimizing large-scale ML training and inference systems, currently at Isomorphic Labs. He has driven practical improvements across PyTorch and Hugging Face Transformers—enabling robust checkpointing and optimizer scheduling that streamline fine-tuning workflows—and has interned on core performance teams at NVIDIA and Databricks working on GPU memory, kernel autotuning, and multi-thousand-GPU training runs. His background includes large-scale optimizer research with Google Brain and hands-on LLM training and inference engineering at Cohere, giving him deep fluency in CUDA, NCCL, distributed parallelism, and productionizing model stacks. Based in Canada and trained at the University of Waterloo, Bilal mixes research rigor with production pragmatism and prefers to be contacted via bilal2vec.com or bilal2vec@gmail.com.
9 years of coding experience
2 years of employment as a software developer
Bachelor of Software Engineering, Bachelor of Software Engineering at University of Waterloo
English
Github Skills (14)
transformers10
pytorch10
machine-learning10
deeplearning-ai10
nlp10
deep-learning10
checkpoint10
python10
natural-language-processing10
fine-tuning10
adam10
checkpointing10
hub9
language-model9
Programming languages (9)
Github contributions (5)
huggingface/transformers
Oct 2019 - Dec 2019
🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
Role in this project:
ML Engineer
ML EngineerContributions:12 commits, 4 PRs, 26 comments in 2 months
Contributions summary:Bilal primarily focused on modifying existing training scripts within the Hugging Face Transformers library. Their contributions centered around enhancing the training process, specifically by enabling the saving and loading of optimizer and scheduler states, and allowing training to resume from saved checkpoints. They also made minor improvements to documentation. These changes streamline the fine-tuning process for various language models within the library.
audioinferencemachine-learning-modelsmultimodaltransformers
bilal2vec/dotfiles
Dec 2020 - Apr 2025
Contributions:55 pushes, 1 branch in 4 years 4 months
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.