Di Jin
Co-Founder at Eigen AI
California, United States
Summary
🤩
Rockstar🎓
Top SchoolDi Jin is a Co-Founder and PhD-trained AI researcher with nine years of experience and a decade-long focus on NLP, having authored 60+ papers with 5k+ citations. He has led model alignment and post-training efforts for LLMs at Meta and Amazon, inventing practical methods for aligning large models to human preferences (e.g., reward-based data augmentation and constitutional critique). At Eigen AI he now applies that expertise to make enterprise AI leaner through post-training, compression, and optimization. His contributions to open-source NLP tooling include adding a Particle Swarm Optimization adversarial search to the widely used TextAttack framework, reflecting a blend of theoretical rigor and production engineering. Based in California, he combines top-tier academic training from MIT and Tsinghua with hands-on impact across research, product, and open-source communities.
9 years of coding experience
10 years of employment as a software developer
Doctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at Massachusetts Institute of TechnologyBachelor's degree Precision Instruments, Bachelor's degree Precision Instruments at Tsinghua University
Github Skills (7)
algorithm10
algorithms10
machine-learning10
nlp10
python10
pytorch9
data-augmentation7
Programming languages (5)
Github contributions (5)
QData/TextAttack
Jun 2020 - Jul 2020
TextAttack 🐙 is a Python framework for adversarial attacks, data augmentation, and model training in NLP https://textattack.readthedocs.io/en/master/
Role in this project:
ML Engineer
ML EngineerContributions:18 commits, 2 PRs, 11 pushes in 6 days
Contributions summary:Di implemented a Particle Swarm Optimization (PSO) algorithm for adversarial attacks, integrating it as a new search method within the TextAttack framework. They added a PSO model, including a class and associated functions for the algorithm. The user's changes involved modifications to the core search method logic and integration with other components of the TextAttack library. The user also updated the HowNet transformation to include an assert statement, ensuring that the number of words are identical before running.
pythonadversarial-machine-learningadversarialmodel-trainingtraining
jind11/MedQA
Sep 2020 - Dec 2022
Code and data for MedQA
Contributions:6 commits, 2 PRs, 7 pushes in 2 years 2 months
python
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.