I finished my PhD at Brown University (advised by Stephen Bach), where I was supported by the Open Philanthropy (now Coefficient Giving) grant for technical AI safety.
I was recently an Astra Safety Research Fellow (mentored by Miles Wang and Olivia Watkins) and will be joining OpenAI soon.
Research
I work on AI safety. I am most interested in the science of misalignment and loss-of-control (e.g., how they arise during training and deployment), as well as the implications and limitations of recursively self-improving AI.
I currently work on model organisms and scalable oversight, such as CoT monitorability. My past work includes:
- Frontier risk and preparedness evaluations, such as for Kimi K2.5 (preprint 2026).
- Generalization of alignment, studied via mech interp (EMNLP 2024; ICLR 2026).
- Adversarial robustness to (multilingual) jailbreaks (Best Paper @ NeurIPS 2023 SoLaR; NAACL 2025).
I previously worked on multilingual LLMs so that AI systems can benefit everyone equitably and safely. Some previous work: ACL 2023; Best Paper @ ACL 2024; INTERSPEECH 2025.
Fun Facts
- I went to Minerva University for my undergrad and had the opportunity to live in 6 different cities (SF, Seoul, Hyderabad, Berlin, Buenos Aires, and London) during those four years.
- I was really into biomedical research before I got into AI. My first research project was to create a breathalyzer for lung cancer.
- I love dancing. I have been dancing salsa/bachata for a while, and have picked up a bit of Lindy Hop and tango in my journey.