Research Scientist, AI Controls and Monitoring
Core
Design methods, systems, and experiments to ensure advanced AI models and agents remain aligned with intended goals in high-stakes or adversarial environments.
Role type
Research Scientist (AI Safety & Control)
Builds
Monitoring techniques, observability methods, layered control mechanisms, fail-safes, oversight protocols, and red-team simulations.
Domain
Artificial Intelligence Safety, AI Alignment, Policy Research
Deliverable
production ML models
Required skills
Machine learning research, generative AI expertise, experimental design, prototyping, system debugging, cross-functional collaboration
Preferred skills
Runtime monitoring, anomaly detection, ML observability, AI control/alignment research (scalable oversight, interpretability, debate), post-training techniques (RLHF, DPO, GRPO)
Technologies
RLHF, DPO, GRPO, red-team simulation frameworks
Responsibilities
Develop real-time monitoring techniques to track AI behavior and flag deviations; Research mechanisms for layered control including fail-safes and intervention methods; Design red-team simulations to probe weaknesses in oversight; Collaborate with policymakers and engineers to establish standards and benchmarks.
Seniority
Mid-Senior, hands-on IC
