Cbai The Research Manager, AI Safety at Cbai supports research fellows and mentors working on AI safety topics including interpretability, AI control, and governance. This role involves research coaching, program development, and community building.
Responsibilities
You'll work closely with research fellows and their mentors — renowned researchers from Cambridge and beyond — to support cutting-edge work on interpretability, AI control, formal verification for provably safe AI, evaluations, and various topics in AI governance & policy. We're interested in hiring research managers with technical research experience and governance & policy research experience.
Research Management (0.7 FTE)
Conduct frequent 1-1s with your fellows, providing feedback on research progress and helping them overcome obstacles, coaching them through challenges such as but not limited to debugging the research and preparing literature scaffolds, and supporting data collection & analysis, as well as methodology development for the experiments & hypothesis tests
Provide feedback on your fellows' research and help CBAI create an environment that nudges your fellows to be more rigorous in their approach
Connect fellows with resources, literature, and opportunities they can explore during and after the fellowship program
Communicate with your fellows' mentors to define clear research objectives and support your fellows’ research progression
Contribute to the fellow selection process by reviewing and interviewing candidates to ensure the cohort consists of individuals with strong epistemic standards
Program Management & Development (0.3 FTE)
Design reading group curriculum components and workshop programs
Curate a speaker event series based on fellow profiles and recent important studies
Support special projects aligned with your strengths (e.g., applicant selection, evaluation frameworks, mentor onboarding)
Meet weekly with program leadership to enhance feedback loops and continuously improve the program
Qualification
Excellent communicatorInterpersonally excellentMission-motivatedOrganized and conscientiousCurious and adaptablePublished research in interpretabilityHoweverWhy This Role May Not Be the Right FitYou're supporting others' research
Required
We expect you to be characterized by most of the qualities listed below.
Experience supporting complex intellectual work. You have helped others execute complex analytical or research projects. This might come from teaching, managing technical teams, conducting your own research, consulting, coordinating academic programs, or providing substantive feedback on others' work.
You understand the research process from the inside. You've done substantial analytical work yourself (whether in academia, policy analysis, consulting, or industry research). This means you can recognize when a research plan is solid vs. hand-wavy, identify blockers in someone's thinking, and suggest concrete next steps.
You're skilled at developmental feedback. You provide constructive feedback that advances the work, rather than simply identifying problems. You can help people strengthen their arguments, tighten their methodology, and present their findings more clearly, even when the specific topic is outside your core expertise.
Excellent communicator. You explain complex concepts clearly and give constructive feedback effectively. You check in proactively when you're unsure about something or notice a potential problem.
Interpersonally excellent. You have genuine empathy, active listening skills, and a servant leadership approach. You enjoy helping others succeed and take pride in their accomplishments. You're skilled at managing stakeholders and coordinating between multiple parties (researchers, advisors, administrators).
Mission-motivated. You are strongly aligned with CBAI's mission and are familiar with AGI safety and catastrophic AI risks. You want to contribute meaningfully to reducing AI catastrophic risks and are passionate about accomplishing as much as possible.
Organized and conscientious. You keep complex projects on track, follow through reliably, and maintain clear communication. You're receptive to feedback and continuously improve your approach based on what you learn.
Curious and adaptable. You're excited to learn about diverse research agendas and work with researchers from varied backgrounds. You actively seek out tools and approaches that can improve fellowship effectiveness.
The ideal candidate for this position will possess most of the qualities described above and will have a bachelor's degree or higher in Computer Science, Mathematics, Statistics, Economics, Public Policy, Political Science, or a related field, with research experience demonstrating strong methodological knowledge and a genuine interest in building a career in AI safety research.
Preferred
Previous involvement in AI safety/alignment programs or similar field-building initiatives
Published research in interpretability, AI control, or adjacent agendas
Experience managing research programs or academic initiatives
However, there is no such thing as a "perfect" candidate. If you are on the fence about applying because you are unsure whether you are qualified, we strongly encourage you to apply.
Why This Role May Not Be the Right Fit
We want to be transparent about what this position entails so you can make an informed decision about whether it's right for you:
You're supporting others' research. Your primary job is helping fellows execute their research projects, not shaping their research strategy. You'll coach fellows through challenges, provide feedback, and connect them with resources, but you won't set research directions or objectives; mentors and fellows own that. In some cases, with mentor and fellow approval, you may contribute substantially enough to merit co-authorship, but this is a facilitation role first and foremost.
Your success is measured by fellows' accomplishments. When fellows publish papers, secure positions at top labs, or advance in their careers, that's your win. If you're primarily motivated by building your own research reputation or publication record, this role won't be satisfying.
