OpenAI
Researcher, Alignment CoT Monitorability
$295K–$500K
San Francisco · Full-time · Posted Aug 4, 2026
About the job
Research chain-of-thought monitorability and develop experiments, evaluations, and monitoring approaches that support scalable oversight and safer model training.
What you'll do
- Design empirical studies and evaluations of monitorability across reasoning models and training settings.
- Investigate how training interventions affect monitorability and turn findings into oversight and training recommendations.
- Analyze model behavior, develop monitoring methods, and collaborate on applying them to training runs.
What they ask for
- Strong hands-on experience training, evaluating, or debugging large ML models, especially LLMs.
- Depth in alignment, interpretability, model behavior, empirical ML, or adjacent research.
- Ability to turn ambiguous research questions into measurable experiments and follow evidence.
- Ability to combine research ideation with engineering execution.
Nice to have
- Direct chain-of-thought interpretability experience.
- Experience in broader interpretability, alignment, model training, or investigative model-behavior work.
Pay
$295K – $500K USD base salary
Tools
LLMs
AI focus
Research chain-of-thought and broader model monitorability, investigate how training affects legibility, and improve monitoring and scalable oversight for alignment and safety.
Model-written summary · the original posting is authoritative