Sotabase

Career

· AI Researcher, Google DeepMind2024–
· PhD Student, Stanford Artificial Intelligence Laboratory (SAIL)2023–
· PhD in Computer Science, Stanford University2023–
· Software Engineer, Messenger, Meta2019–2020
· Software Engineer Intern, Safety Engineering, Amazon Web Services2018–2019
· Software Engineer Intern, Core OS, Apple2018–
· Teaching Assistant, The University of Sydney2017–2020
· MS Research Student, Carnegie Mellon University
· AI Resident, Google Research
· AI Resident Researcher, Stanford University
· Research Assistant (RI/MLD), Stanford University
· Research & Teaching Assistant, Stanford University

Publications (20)

Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition
Computer Vision and Pattern Recognition · 2020
998
cited
The Distributed Discrete Gaussian Mechanism for Federated Learning with Secure Aggregation
International Conference on Machine Learning · 2021
280
cited
The Skellam Mechanism for Differentially Private Federated Learning
Neural Information Processing Systems · 2021
148
cited
Learning Implicit Credit Assignment for Cooperative Multi-Agent Reinforcement Learning
Neural Information Processing Systems · 2020
121
cited
Investigating Data Contamination for Pre-training Language Models
arXiv.org · 2024
93
cited
On Privacy and Personalization in Cross-Silo Federated Learning
Neural Information Processing Systems · 2022
70
cited
Motley: Benchmarking Heterogeneity and Personalization in Federated Learning
arXiv.org · 2022
53
cited
Learning Implicit Credit Assignment for Multi-Agent Actor-Critic
arXiv.org · 2020
28
cited
Differentially Private Adaptive Optimization with Delayed Preconditioners
International Conference on Learning Representations · 2022
14
cited
On Fairness of Low-Rank Adaptation of Large Models
arXiv.org · 2024
7
cited
UQ: Assessing Language Models on Unsolved Questions
arXiv.org · 2025
3
cited
Understanding Adversarial Transfer: Why Representation-Space Attacks Fail Where Data-Space Attacks Succeed
arXiv.org · 2025
1
cited
D OES D ATA C ONTAMINATION M AKE A D IFFERENCE ? I NSIGHTS FROM I NTENTIONALLY C ONTAMINATING P RE - TRAINING D ATA F OR L ANGUAGE M ODELS
O N F AIRNESS I MPLICATIONS AND E VALUATIONS OF L OW -R ANK A DAPTATION OF L ARGE M ODELS
Quantifying the Effect of Test Set Contamination on Generative Evaluations
2026
Sotabase