Senior Research Scientist

About the role

Join Faculty's central R&D team to lead the creation of advanced safety evaluations for AI systems. This role involves conducting original research to improve AI safety methods and collaborating with teams that perform evaluations and red-teaming for leading AI labs. You will produce scientific outputs that influence the future of safe AI deployment in real-world applications.

Key facts

What you'll do

• Lead the creation of new safety evaluations in critical areas like CBRN and Cyber to measure emerging AI risks.
• Conduct original technical research in AI safety evaluation methods, guiding projects from initial concept to publication.
• Help shape the R&D strategy by finding opportunities to improve safety evaluation methods across Faculty and the broader AI community.
• Provide expert technical knowledge and thought leadership to client projects, evaluation efforts, and red-teaming for frontier AI labs.
• Represent Faculty's scientific leadership through active participation with the global research community, leading AI labs, and government bodies.

Requirements

• A proven history of managing research projects from problem identification to publication, driven by scientific curiosity.
• Practical experience designing and building AI evaluations or benchmarks, with a strong understanding of construct validity and how to avoid confounding factors.
• Expertise in red-teaming, adversarial testing, jailbreaking, indirect prompt injection, and assessing the robustness of model safeguards.
• Strong foundational skills in experimental design, statistical analysis, and methods for quantifying uncertainty, including Bayesian approaches.
• Deep understanding of language models, generative AI architectures, training methods, and safety mitigation techniques.
• Solid proficiency in Python and the engineering discipline needed to produce reliable and reproducible research.

Nice to have

• Experience with threat and risk modeling.
• Background or knowledge in high-risk fields such as CBRN or Cybersecurity.
• A history of impactful AI research, demonstrated by top-tier publications or equivalent practical achievements.

Skills & tools

• Python
• Experimental design
• Statistical analysis
• Uncertainty quantification
• Bayesian methods

Practical notes

• Benefits include unlimited annual leave, private healthcare and dental, enhanced parental leave, family-friendly flexibility, Sanctus Coaching, and hybrid working.
• Faculty encourages applications from diverse backgrounds and is open to discussing part-time hours.