Research
My interests run wider than my published work, and I’m always glad to hear from people working on adjacent problems. Three streams cover most of what I’m doing now.
Compute and hardware governance
I examine technical verification mechanisms for international access frameworks and the trade-offs created by open-weight model release. Big things cooking here that will be announced soon!
AI-enabled information manipulation
The subject of my thesis at Yale where I’m using multilingual natural language processing to identify coordinated inauthentic behavior in Taiwan. The project culminates in an October research visit to Taipei with Doublethink Lab and Taiwan AI Labs. It also shapes the SPAR project I’m co-mentoring this fall on Taiwan’s position under transformative AI.
Evaluation and auditing of frontier models in agentic settings
I’m building methods for evaluating frontier models in agentic settings, including the limits of post-hoc log review as a safeguard and how model behavior shifts across languages and extended interaction horizons. This is my work at GovAI this summer, which I’m writing up for a NeurIPS workshop.
Across all three streams I have grown more concerned about epistemic security than about the chemical and biological risk pathways that dominate current frontier safety debates. If you’re working on epistemic risk and keeping human truth in AI, please reach out.
Current affiliations
| Organization | Role |
|---|---|
| Centre for the Governance of AI (GovAI) | Summer Research Fellow |
| OpenAI | Research Collaborator |
| Council on Foreign Relations | Graduate Research Assistant |
| Doublethink Lab | Research Collaborator |
| California Governor’s Office of Emergency Services | Research Collaborator |
| Supervised Program for Alignment Research (SPAR) | Co-Mentor, Fall 2026 |
| Yale University | Graduate Teaching Fellow |