Abhineet Agarwal
Member Of Technical Staff at Traversal
Berkeley, California, United States
Summary
🤩
Rockstar🎓
Top SchoolAbhineet Agarwal is a Member of Technical Staff based in Berkeley with a PhD in Statistics from UC Berkeley focused on AI explainability, causal inference, and reinforcement learning. He builds AI agents for DevOps observability at Traversal and brings prior quant research experience from an internship at Citadel where he worked on equity alpha generation. His doctoral work and projects emphasize interpretable models and decision-making under uncertainty, reflected in open-source contributions to the imodels library improving tree-based interpretability and efficient approximate LOO cross-validation. With a background in physics and math from Columbia and early research on quantum optics simulations at the Flatiron Institute, he combines strong theoretical rigor with practical engineering. Colleagues describe him as someone who bridges cutting-edge ML research and production-grade tooling, often optimizing classical algorithms for modern ML stacks.
5 years of coding experience
Doctor of Philosophy - PhD, Statistics, 3.93, Doctor of Philosophy - PhD, Statistics, 3.93 at University of California, BerkeleyBachelor of Arts - BA, Physics and Mathematics, 3.94, Bachelor of Arts - BA, Physics and Mathematics, 3.94 at Columbia University in the City of New York
Github Skills (10)
scikit10
scikit-learn10
decision-tree10
machine-learning10
interpretation10
python10
data-science10
gradient-boosting9
artificial-intelligence7
explainable-artificial-intelligence7
Programming languages (3)
Github contributions (5)
csinva/imodels
Dec 2021 - Nov 2022
Interpretable ML package 🔍 for concise, transparent, and accurate predictive modeling (sklearn-compatible).
Role in this project:
Data Scientist
Data ScientistContributions:68 commits, 69 pushes, 1 comment in 10 months
Contributions summary:Abhineet primarily contributed to the `imodels` repository by implementing and improving tree-based interpretable machine learning models. Their work involved modifications to the `ShrunkTree` class, including the addition of different shrinkage schemes and bug fixes to ensure compatibility with various machine learning methods, specifically Gradient Boosting. Furthermore, the user integrated shrinkage techniques within CART with cost complexity pruning (CCP) for both classification and regression tasks, and added the approximate leave one out cross validation to reduce computation time.
pythonxaisklearnsklearn-compatibleinterpretable
aagarwal1996/PCS_UQ
Mar 2025 - Mar 2025
Repository for generating prediction intervals via the predictability-computability-stability (PCS) framework.
Contributions:72 pushes, 1 branch in 22 days
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.