Product Manager, Safety Measurement
San Francisco • FullTime
Posted 5mo ago
Remote Work Policy
On-site
Employment Type
FullTime
Categories
Applied AI Engineer
About the job
The Safety Measurement Product Manager will lead OpenAI's strategy for measuring harm and the effectiveness of safeguards in production. This role involves driving the development and strategy for a suite of safety measurement platforms and products used across the company. You will collaborate closely with safety research and engineering teams to define what, where, and how we measure safety, influencing critical leadership decisions and informing ongoing safety initiatives. You will also represent the company's primary safety metric and prioritize requests to expand the safety measurement platform to new use cases.
Responsibilities
- Partner with data science, research, engineering, policy, and other stakeholders to define a vision for understanding safety outcomes and prevalence on our platforms.
- Define strategic priorities and product roadmaps for improving safety measurement approaches and scaling the measurement platform.
- Establish repeatable processes to integrate AI safety research into OpenAI's safety measurement products.
- Develop and refine clear, actionable success criteria for safety measurement.
- Represent quantitative progress on safety to senior leadership.
Requirements
- 6+ years of product management or related industry experience, with expertise in AI safety, trust & safety, integrity, or similar domains.
- Background in data science, statistics, economics, or related fields.
- Deep curiosity and interest in interdisciplinary fields like human-computer interaction, psychology, or philosophy.
- Hands-on experience driving consensus and action in ambiguous environments.
- Ability to identify and challenge underlying assumptions and constraints through questioning.
- Highly effective at cross-functional collaboration and communicating complex technical concepts clearly.
- Enjoy working in a fast-paced, high-growth environment.