Technical teamVolunteer role
Research Mentor, Technical AI Safety
Mentor top graduates on multi-agent safety, interpretability and automated red-teaming projects.
- 3 to 5Hours a week
- Ongoing mentorshipFormat
- RemoteWhere you work
- TechnicalTeam
The role
You mentor the research of our top cohort graduates. The main focus is multi-agent safety, and we also want mentors from interpretability, evals and red-teaming. Your job is to help researchers scope projects, review drafts and get the work out into the world.
What you will do
- Mentor 2 to 4 researchers per cycle on technical AI safety projects
- Focus on multi-agent safety: emergent coordination, collusion, cascade failures
- Take on interpretability, evals or red-teaming projects as well
- Review drafts and push them towards publication: papers, EA Forum, LessWrong, Substack
- Aim for 3 to 10 research outputs per cycle across your mentees
What we look for
- An active research background in multi-agent systems, interpretability, evals or red-teaming
- At least one published paper, preprint, workshop paper or substantial EA Forum post
- Ability to give honest, structured feedback on research drafts
- 3 to 5 hours a week, consistently
Recommended background: Published researchers, PhD candidates, SPAR, MATS or ARENA alumni, or senior alignment practitioners.
What you get
Co-authorship on research papers
A letter of recommendation for your own fellowship or role applications
Priority for paid roles when funding arrives
How selection works
- 1
Application form
Your background, any AI safety courses or projects, and your weekly time, in the form below.
- 2
Work test
A short, practical task for your role, done in your own time, so we can see how you work.
- 3
Interview
A conversation with our core team about fit, cohort logistics, mentorship support and onboarding.
Apply for this role
Two short steps, for this role only.
About the role
Your time, and your experience for this role.
A paper, preprint, workshop paper or substantial EA Forum post.
About you
Your background, and what you have done in AI safety.
Whichever shows your research best.
Study, work, and the tools you use.
BlueDot, reading groups, research, posts. Leave it empty if none yet.
Application received, there.
Your application for Research Mentor, Technical AI Safety is in. A confirmation is on its way to your email.
- 01We read itEvery application, by our core team.
- 02A work testA short task for your role, in your own time.
- 03An interviewA conversation about fit and onboarding.
Reference VOL