Research Scientist (Control)
Core
Design and implement control protocols and monitoring systems for AI coding agents to detect security vulnerabilities and failure modes at scale.
Role type
Research Scientist (AI Control & Safety)
Builds
Hierarchical monitoring systems, fine-tuned open-source models, and evaluation frameworks for AI agent safety.
Domain
Artificial Intelligence, AI Safety, Software Engineering
Deliverable
production ML models
Required skills
Empirical research with LLMs, AI coding agents, LLM-as-a-judge setups, experimental design, Python programming, independent research
Preferred skills
AI safety concepts, computer security, fine-tuning language models, developer tools, production log analysis
Technologies
Python, LLMs, Open-source models
Responsibilities
Design and conduct experiments to test monitor effectiveness; Build and maintain evaluation frameworks and high-quality datasets; Fine-tune open-source models to create efficient monitors; Design agentic monitoring systems to investigate logs; Contribute discovered failure modes to shared libraries; Stay current with AI safety and coding security research.