Postdoctoral Associate, AI Performance and Predictability
MIT FutureTechCambridge, MA
AI safety & policy
Full-time
AI safety & policy
Full-time
Routes to impact
Skill-building & building career capital
Direct high impact on an important cause
Description
Conduct empirical AI research on performance limits, safety, and predictability in advanced systems, designing experiments and evaluation tools for deep learning and multi-agent settings.
- Design and analyze experiments on scaling, robustness, safety, and generalization
- Develop agentful arenas to evaluate AI agents and modifications
- Build tools and frameworks for assessing AI alignment techniques
- Publish research papers, white papers, and public reports; contribute to external workshops and collaborations
- Starting in 2027 for 1 year, renewable
This text was generated by AI. If you notice any inconsistencies, please let us know using this form.



