Thomas Fan
Member Of Technical Staff at scikit-learn
New York, New York, United States
Summary
🤩
Rockstar🎓
Top SchoolThomas Fan is a machine learning-focused software engineer with 11 years of experience and 6+ years of active open-source stewardship, currently a Member of Technical Staff at Modal and a core developer and Technical Committee member for scikit-learn. He combines deep algorithmic and systems skills—shipping features like Array API and GPU support for scikit-learn and contributing to heavy-hitting projects such as NumPy, SciPy, PyTorch, and LightGBM—with a strong focus on testing, CI/CD, and developer experience. His background spans ML research engineering, production model serving, and DevOps, including building LLM serving infrastructure and optimizing container build systems. A prolific speaker, he regularly presents on performance, parallelism, and practical ML tooling at SciPy, PyData, and ODSC, and has driven impactful contributions like Cython optimizations and histogram-based gradient boosting improvements. He blends physics and math training with pragmatic engineering, frequently improving testing, documentation, and CI across the scientific Python ecosystem.
11 years of coding experience
9 years of employment as a software developer
Master of Science - MS Mathematics, Master of Science - MS Mathematics at New York University - Polytechnic School of EngineeringMaster of Arts - MA Physics, Master of Arts - MA Physics at Stony Brook University
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (84)
probabilistic-programming10
python10
testing10
datasets10
mcmc10
build-automation10
haproxy10
scientific-computing10
bayesian-inference10
plasma-physics10
scikit10
pandas10
dockers10
numpy10
compiler10
Programming languages (19)
Github contributions (5)
scikit-learn/scikit-learn
Jun 2018 - Jan 2023
scikit-learn: machine learning in Python
Role in this project:
Data Scientist & ML Engineer
Data Scientist & ML EngineerContributions:5407 reviews, 984 commits, 2538 PRs in 4 years 8 months
Contributions summary:Thomas contributed to scikit-learn by addressing a bug in `tree.export_text` related to single feature names, upgrading the CI to PyPy 7.1.1, adding support for pandas DataFrame in OpenML and improving error handling, and fixing an issue that affected the multi-class ROC AUC scorer. They also worked on feature addition to and optimization of, among other modules, the core of scikit-learn such as the `ensemble` module.
data-analysispythonstatisticsdata-sciencelearn-machine-learning
skorch-dev/skorch
Mar 2018 - Jul 2021
A scikit-learn compatible neural network library that wraps PyTorch
Role in this project:ML Engineer
ML EngineerContributions:212 reviews, 97 commits, 108 PRs in 3 years 4 months
Contributions summary:Thomas implemented and integrated a new Cyclic Learning Rate (CLR) scheduler to the skorch library, enhancing its machine learning capabilities. They also addressed target shape issues within the NeuralNetClassifier, ensuring compatibility, and refactored code to add support for PyTorch 0.4. Additionally, they contributed to testing procedures, modified the tolerance level, and added a test to verify the learning rate scheduler. They implemented the history loading and saving functionality.
pytorchpythonwrapsneural-networksmachine-learning
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.