Yichen Xu
PhD Candidate
Wuxi City, Jiangsu, United States
Summary
🤩
Rockstar🎓
Top SchoolYichen Xu is a PhD candidate in Biostatistics at UC Berkeley with nine years of software engineering experience bridging causal inference research and practical ML systems. He contributes to high-impact open-source projects, from implementing GADT and type-system fixes in the Scala 3 compiler to developing graph contrastive learning augmentations and loss functions in PyGCL for molecule property prediction. His work blends deep statistical training with hands-on ML engineering—moving ideas like TMLE and advanced augmentations into reproducible code and training pipelines. Based in Wuxi but active in international research and open-source communities, he brings a rare combination of formal math, compiler-level rigor, and applied graph ML expertise.
9 years of coding experience
MA, PhD Student, Biostatistics, MA, PhD Student, Biostatistics at University of California, BerkeleyBachelor of Science - BS, Applied Mathmatics, Bachelor of Science - BS, Applied Mathmatics at The Chinese University of Hong Kong, Shenzhen 香港中文大学（深圳）
Github Skills (16)
type-system10
pytorch10
machine-learning10
compiler-design10
pattern-matching10
python10
scala10
model-testing9
evaluation9
eval9
test-integration8
unit-testing8
integration-testing8
testing8
unit-test8
Programming languages (13)
Github contributions (5)
PyGCL/PyGCL
Jun 2021 - Apr 2022
PyGCL: A PyTorch Library for Graph Contrastive Learning
Role in this project:
ML Engineer
ML EngineerContributions:7 reviews, 109 commits, 22 PRs in 10 months
Contributions summary:Yichen contributed code related to graph contrastive learning, the core focus of the repository. Their commits included the addition of new augmentation techniques like EdgeAttrMasking and EdgeAttrDropout, as well as the implementation of various loss functions, including Barlow Twins and VICReg, tailored for graph-based contrastive learning. They also updated training scripts for different graph datasets and model architectures (e.g., GINConv), adapting existing code for tasks like molecule property prediction and graph classification, and added a trial-based training framework.
pytorchcontrastive-learningmachine-learningcontrastivegraph-representation-learning
scala/scala3
Feb 2021 - Jan 2023
The Scala 3 compiler, also known as Dotty.
Role in this project:Back-end Developer
Back-end DeveloperContributions:124 reviews, 96 commits, 118 PRs in 1 year 10 months
Contributions summary:Yichen primarily contributed to the Scala 3 compiler, focusing on pattern matching, type system improvements, and GADT support. Their work involved fixing bugs related to refinement types, GADTs, and handling of capture sets, and also included the addition of new test cases to validate the improvements. Furthermore, the user made contributions to bounds propagation, particularly in the context of constraint handling and type parameter unification.
compilerscala3scaladottyepfl
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.