I do research on AI safety and security. Reach out to me via email if you’d like to get in contact for discussion or collaboration. I’m always open to talk about research ideas.

I’m broadly interested in the security and robustness for deep learning models as well as understanding their capabilities for generalization in out of distribution settings. Right now, I’m particularly focused on the ability of LLMs to encode and interpret subliminal information.

At NYU, I’m advised by Rico Angell and He He, working on jailbreak defenses and model interpretability. I’m also extremely fortunate to be mentored by Chawin Sitawarin during my undergraduate studies at UC Berkeley as a part of David Wagner’s group.

In my free time, I enjoy reading and writing fantasy, as well as weightlifting.

I’m also a fan of landscape photography. Here’s a randomly sampled photo from my portfolio:

Publications

Stronger Universal and Transfer Attacks by Suppressing Refusals
Huang, D., Shah, A., Araujo, A., Wagner, D., & Sitawarin, C.
NAACL 2025 (abridged version accepted NeurIPS SafeGenAI 2024), 2025
A novel algorithm leveraging model refusal representation for automated jailbreaking suffix generation on LLMs

Efficient Mitigation of Bus Bunching through Setter-Based Curriculum Learning
Avidan Shah, Danny Tran, Yuhan Tang
arxiv, 2023
This paper serves as the final project for CS285, and was not submitted to any conference for review