Loss of Control Lead (Frontier Safety)
Мэтч & Сопровод
Для мэтча с этой вакансией нужен Plus
Описание вакансии
Location: Remote globally; some timezone flexibility is expected. Full-time, 40 hours per week. If based in the USA or Singapore, employment is direct; outside those locations, employment is through an EOR or as a contractor.
Salary: $225,000–$450,000/year depending on experience and location, with potential additional compensation for exceptional candidates.
Company
is a non-profit AI research institute focused on AI safety, frontier-system evaluations, risk research, and coordinated responses to advanced AI risks.
What you will do
- Build and lead a high-impact loss-of-control testing program with frontier AI companies and governments.
- Develop evaluations, testing methodologies, playbooks, benchmarks, and analysis tools that identify unsafe behavior and safeguard gaps.
- Plan and execute an evaluation portfolio from early model training through post-deployment and from inference-only access to deeply embedded evaluations.
- Produce private stakeholder reports and support public reports, leaderboards, and standards that improve frontier AI safeguards.
- Manage, mentor, recruit, and grow an elite technical team while maintaining high standards for velocity, rigor, judgment, and impact.
- Collaborate across and the Frontier Safety division to turn technical findings into real-world safety outcomes.
Requirements
- Strong scientific track record in AI or another highly technical field such as computer science, mathematics, or physics.
- Strong record of building evaluations, benchmarks, datasets, or experiments testing AI capabilities or behavior.
- Strong record of work and impact in loss-of-control or closely related risks.
- Strong record of managing, growing, and leading technical teams and leading strategic technical programs across multiple projects.
- Ability to thrive in rapidly evolving, ambiguous environments and demonstrate drive for mission and impact.
- Ability to communicate technical ideas effectively to technical and non-technical audiences.
- Demonstrated relentlessness in achieving ambitious goals and willingness to execute hands-on work rather than operate only at the strategy level.
- Ability to work collaboratively on technical direction rather than setting it unilaterally.
- Deep domain expertise or a demonstrated ability to build exceptional programs by coordinating others with that expertise.
Nice to have
- Experience building technical products in an entrepreneurial environment.
- Ability to discover non-obvious, high-severity risks or vulnerabilities in complex systems.
- Experience in alignment, cybersecurity, red-teaming, or other closely related areas.
- Previous collaboration with AI labs or government safety institutes.
- Published work in AI safety, control, evaluation, security, or robustness.
Culture & Benefits
- Mission-driven, high-impact work focused on improving the safety and security of advanced AI systems.
- Fast-moving environment with evolving threat models and opportunities to build new technical programs.
- Health insurance with 94% of the premium covered for eligible full-time US employees.
- 401(k) contribution match of up to 2%, 25 days of paid time off, and up to 10 sick days for eligible full-time US employees.
- Paid bereavement, family, medical, and pregnancy disability leave, plus a work-from-home stipend and equipment.
- Work-related travel and equipment expenses are covered; catered meals are available at the Berkeley office.
Hiring process
- Submit a CV and answers to questions about interest in the role through the application form; do not email a resume separately.
- Process: short screening call, work test, interviews, and a final paid work trial. Alternative fit-testing may be available if a work trial is not possible.
Будьте осторожны
Если работодатель просит войти в их систему, используя iCloud/Google, прислать код/пароль, запустить код/ПО, не делайте этого. Это мошенники. Обязательно жмите «Пожаловаться» или пишите в поддержку. Подробнее в гайде →