The opportunity
Safety Systems manages the complete lifecycle of safety efforts for OpenAI’s frontier models, ensuring our models are deployed responsibly and have a positive impact on society. Our work spans diverse research and engineering initiatives—from system-level safeguards and model…
What you'll do
Define strategic priorities and product roadmaps to improve safety: measurement and evaluation approaches and scale our measurement platform to new use cases, products, and the needs of cross-functional teams
Partner closely with research, data science, engineering, policy teams, and: other stakeholders to craft a vision for understanding safety outcomes and prevalence across our products, and translate that vision into concrete programs of work.
Drive the integration of cutting-edge AI safety research into OpenAI’s safety: measurement and evaluation platforms, working with researchers, data scientists, and engineers to navigate technical tradeoffs, coordinate dependencies, and resolve blockers.
Develop and continuously refine clear, actionable success criteria that: effectively capture our ambitions with safety measurement and evals.
Represent quantitative progress on safety to senior leadership.
Have 6+ years of experience in research program management, technical program: management, product management, or related roles, with expertise in AI safety, trust & safety, integrity, or related domains.
What they're looking for
- Have a strong foundation in data science, statistics, economics, or a related: field, and hands-on experience with measurement or evaluations.
- Enjoy working at the intersection of data science and AI safety research,: translating evolving measurement needs into clear requirements, priorities, and roadmaps.
- Can independently drive complex, cross-functional programs from problem: definition through rollout, building alignment and unblocking decisions in ambiguous spaces.
- Ask incisive questions, challenge assumptions, and exercise sound judgment: about technical tradeoffs and the strengths and limitations of metrics and evaluations.
