Amirhossein Kazemnejad
Research Assistant at Mila - Quebec Artificial Intelligence Institute
Montreal, Quebec, Canada
Summary
👤
Senior🎓
Top SchoolAmirhossein Kazemnejad is a research-focused machine learning engineer with a decade of experience bridging foundational RL algorithms and large language model (LLM) tuning. Based at Mila and formerly a McGill graduate student, he developed NoPE—work that influenced recent LLMs like Llama 4—and more recently co-introduced VinePPO to address credit assignment problems in RL-for-LLM fine-tuning. His research spans positional encoding, compositional generalization, knowledge acquisition vs. utilization, and practical tools for RL-for-LLM workflows, including the open-source nano-aha-moment library and lecture series. Known for turning theoretical insights into components adopted by major architectures, he combines rigorous academic collaboration with hands-on implementation. Based in Montreal, he continues to focus on fundamental algorithmic improvements that make LLM tuning more robust and interpretable.
10 years of coding experience
High School Diploma, Mathematics, High School Diploma, Mathematics at Allameh Helli 3Bachelors, Computer Engineering, Bachelors, Computer Engineering at Iran University of Science and TechnologyMaster of Science - MS, Computer Science, Master of Science - MS, Computer Science at McGill University
Stackoverflow
834reputation
65kreached
0answers
6questions
Github Skills (57)
tensorflow-addons10
cuda9
deep-reinforcement-learning8
python8
deep-learning8
reinforcement-learning8
inference8
machine-learning8
mathematics8
html-template8
pytorch8
llm7
coding7
gpu7
tensorflow7
Programming languages (7)
Github contributions (5)
kazemnejad/pt_hf_base
Feb 2022 - Aug 2023
Contributions:153 pushes in 1 year 5 months
McGill-NLP/length-generalization
May 2023 - Aug 2024
Code for the paper "The Impact of Positional Encoding on Length Generalization in Transformers", NeurIPS 2023
Contributions:1 release, 1 PR, 16 pushes in 1 year 3 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.