Important Info
Submit your materials through this Google Form:
We are seeking postdoctoral researchers to develop statistical and computational methods for auditing, monitoring, and intervening in AI systems. The positions are part of the AI Oversight collaboration (aioversight.stanford.edu), an interdisciplinary research collaboration at Stanford University focused on developing a toolbox for the safe deployment of AI systems that are aligned with human values and stated guardrails.
The participating faculty are John Duchi, Ramesh Johari, Vasilis Syrgkanis, Benjamin Van Roy, and Tijana Zrnic. Postdoctoral researchers will collaborate with one or more of these faculty members on projects of mutual interest. The research will draw on methods from statistics, machine learning, reinforcement learning, incentive design and causal inference.
Areas of research focus may include:
- Statistical methods for evaluating and auditing black-box AI systems
- Adaptive monitoring of agentic AI systems during deployment
- Intervention policies and selective human oversight
- Safety contracts that reflect institutional requirements
The project connects foundational research with the development of open-source software for running AI agents with oversight, that incorporates the methods identified in the foundational research in user-friendly and publicly accessible agentic harnesses. Postdoctoral researchers will have opportunities to develop and evaluate methods in software that monitors agent activity, supports intervention and human review, and uses outcomes and feedback to improve oversight. The aim is to provide reusable tools for developers, institutional users, and auditors working with existing AI systems without access to model internals.
Postdoctoral researchers will help define the research agenda and lead theoretical, methodological, and empirical research toward publication in collaboration with the project faculty. Teaching is not required as part of this position.
For inquiries about the positions, contact aioversight@lists.stanford.edu. For more information about the project you can visit aioversight.stanford.edu.
-
A strong mathematical background, including expertise in one or more of statistics, machine learning, reinforcement learning, econometrics, causal inference.
-
A Ph.D. in computer science, statistics, operations research, electrical engineering, economics, or a related field, completed by the start of the appointment
-
Experience with computational methods, research software, and empirical analysis.
-
Prior research experience on topics related to AI safety and prior experience in harness engineering is not required.
-
A cover letter describing your research interests and fit with the project
-
A current CV
-
One representative publication or working paper
-
Contact information for two references