Hanlin Tang
CTO - Neural Networks at Databricks
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolHanlin Tang is a seasoned AI leader and founder with a decade of experience building and scaling machine learning systems, currently serving as CTO for Neural Networks at Databricks after co-founding MosaicML. He led Intel’s AI Lab and helped shape industry benchmarks as a founding member and Power WG Chair of MLPerf Training, with PI roles on multiple DARPA and federal programs. His work spans applied deep reinforcement learning, NLP, and large-model scaling, grounded in a Harvard PhD on recurrent neural networks for human vision and publications in top journals and conferences. A hands-on engineer as well as an executive, he contributes to open-source tooling for model training (notably in MosaicML projects like Composer and LLM-Foundry), improving infrastructure, logging, and CI workflows. Based in San Francisco, he combines academic rigor with startup execution and a track record of moving research into production at hardware and software stack levels.
10 years of coding experience
13 years of employment as a software developer
Taipei American SchoolPhD, Biophysics, PhD, Biophysics at Harvard UniversityB.A., Physics, B.A., Physics at Princeton University
Chinese
Github Skills (13)
neural-network10
pytorch10
machine-learning10
github-ci10
nlp10
deep-learning10
githubaction-workflow10
project-management10
python10
build-automation10
llm10
data-processing9
tensorboard9
Programming languages (3)
Github contributions (5)
mosaicml/composer
Oct 2021 - Oct 2022
Supercharge Your Model Training
Role in this project:
Back-end & DevOps Engineer
Back-end & DevOps EngineerContributions:2 releases, 904 reviews, 132 commits in 1 year
Contributions summary:Hanlin contributed to infrastructure and build process improvements. They added GitHub Actions workflows and issue templates, which streamlines project development and issue tracking. They also addressed code quality by adding license headers to files. The user demonstrated knowledge of project setup and management, improving developer experience.
pytorchml-systemsdeep-learningneural-networksmachine-learning
mosaicml/llm-foundry
Apr 2023 - Mar 2024
LLM training code for Databricks foundation models
Role in this project:ML Engineer
ML EngineerContributions:52 reviews, 22 PRs, 58 pushes in 11 months
Contributions summary:Hanlin primarily contributed to the LLM training code, evidenced by modifications to throughput table generation, fixing model initialization, adding version constraints for dependencies and introducing TensorBoard logger support. They also updated the data processing pipeline by renaming datasets, fixing typing, and refactoring logging. These changes indicate involvement in optimizing and improving the training and logging aspects of the LLM training framework.
deep-learningllmneural-networksnlppytorch
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.