Kexin Huang
PhD Candidate
San Francisco, California, United States
Summary
🤩
RockstarKexin Huang is a PhD candidate and ML-focused software engineer based in San Francisco with nine years of experience building and refining deep learning tools for bioinformatics and drug discovery. At Stanford and in open-source projects like DeepPurpose and Therapeutics Data Commons (TDC), Kexin has driven practical improvements—adding GPU support, fixing dataset and device issues, and refactoring data loaders to modernize dataset hierarchies. Their work spans model implementation, dataset curation, and reproducibility (including a DeepDTA reproduction), demonstrating a blend of rigorous research and production-minded engineering. Notably, Kexin contributes to widely used therapeutic science foundations, signaling both domain expertise in DTI/drug property prediction and a knack for improving the plumbing that makes ML experiments reliable.
9 years of coding experience
Github Skills (16)
bioinformatics10
pandas10
data-loading10
pytorch10
machine-learning10
drug-discovery10
deep-learning10
trainings10
python10
model-testing10
modeling10
data-processing10
datasets9
mask-rcnn9
faster-rcnn9
Programming languages (6)
Github contributions (5)
mims-harvard/TDC
Sep 2020 - Jan 2023
Therapeutics Commons (TDC-2): Multimodal Foundation for Therapeutic Science
Role in this project:
Data Scientist & ML Engineer
Data Scientist & ML EngineerContributions:31 reviews, 537 commits, 163 PRs in 2 years 4 months
Contributions summary:Kexin's contributions primarily involve refactoring and migrating data loading functionalities within the repository. Their work includes modifying the `DrugDataLoader` and associated files to follow an updated hierarchy and dataset naming conventions, indicating a focus on data processing and organization. The user also introduced new code that suggests they are working on datasets related to hERG and potentially utilizing them in experiments related to the code's core domain which is drug discovery.
therapeuticsdata-commonsbioinformaticsprecision-medicinescience
kexinhuang12345/DeepPurpose
Mar 2020 - Mar 2022
A Deep Learning Toolkit for DTI, Drug Property, PPI, DDI, Protein Function Prediction (Bioinformatics)
Role in this project:ML Engineer / Data Scientist
ML Engineer / Data ScientistContributions:12 releases, 262 commits, 46 PRs in 2 years
Contributions summary:Kexin primarily contributed to the implementation and refinement of a deep learning toolkit for drug-target interactions and related bioinformatics tasks. Their work involved fixing errors in the dataset and utility functions, addressing issues with pretraining binary labels, and updating the CNN character configurations. They also tested and updated files related to model training and testing, incorporating improvements such as GPU support and fixing device-related errors. The user successfully integrated the DeepPurpose library and demonstrated the completion of a DeepDTA reproduce.
drug-target-interactionsprotein-functionbioinformaticsdrugchainer
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.