Tianshi C
PHD Student at NVIDIA
Old Toronto, Ontario, Canada
Summary
🤩
Rockstar🎓
Top SchoolTianshi C is a PhD candidate in Computer Science at the University of Toronto and a research assistant at the Vector Institute with 9 years of experience building scalable learning algorithms and generative models. He has industrial research experience as a Deep Learning Research Scientist Intern at NVIDIA, working on privacy-preserving and 3D-aware generative methods, and has contributed core ML improvements to high-profile open-source LLM training code (mosaicml/llm-foundry), notably optimizing attention mechanisms like Grouped Query Attention. His academic work spans few-shot learning, OOD detection in medical imaging, and theoretical analyses published at venues including ICLR and ICML workshops. Combining strong engineering chops with rigorous theory, he brings practical robustness fixes and performance optimizations to large-model training while pursuing cutting-edge generative research.
9 years of coding experience
Bachelor of Applied Science (B.A.Sc.), Engineering Science, 3.97 CGPA,, Bachelor of Applied Science (B.A.Sc.), Engineering Science, 3.97 CGPA, at University of Toronto
English, Chinese, French
Github Skills (11)
neural-network10
attention-mechanism10
pytorch10
machine-learning10
deep-learning10
python10
model-optimization10
llm10
nlp9
triton9
pytest8
Programming languages (3)
Github contributions (5)
mosaicml/llm-foundry
May 2023 - Oct 2024
LLM training code for Databricks foundation models
Role in this project:
ML Engineer
ML EngineerContributions:41 reviews, 28 PRs, 16 pushes in 1 year 5 months
Contributions summary:Tianshi primarily contributed to the core machine learning components of the repository, focused on improving the performance and stability of the language models. Their contributions included refactoring attention mechanisms, specifically implementing and optimizing Grouped Query Attention (GQA) and addressing edge cases in Triton-based attention. Furthermore, the user fixed bugs related to model initialization and padding, which improved the model's robustness. These changes were made to improve the foundation model training and inference capabilities.
deep-learningllmneural-networksnlppytorch
sashaDoubov/llm-foundry
May 2023 - Oct 2024
Contributions:201 pushes, 27 branches in 1 year 5 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.