About the job
As a Data Scientist in Safety Systems, you will establish the data-driven approach for understanding, evaluating, and monitoring the safety of our production systems. You should expect to collaborate with our partners across the company to define north-star metrics, own and implement the statistical methods to productionize those metrics, conduct analysis to understand the impact of our products, and establish source-of-truth dashboards that the entire company can use to answer safety-related questions. Most importantly, you will be a core member of the Safety Systems team, collaborating with researchers and engineers to advance our goals of safe, robust, and reliable AI.
Responsibilities
Lead our efforts in understanding and measuring the real-world safety impacts of OpenAI’s current and upcoming products
Uncovering new ways to improve our approaches to measuring and mitigating harm and abuse
Develop and implement statistical methods necessary to operationalize safety-related metrics
Provide direction, guidance, and coordination of projects in the space
Establish a data-driven culture within Safety Systems by driving the definition, tracking, and operationalizing of feature-, product-, and company-level metrics
Create and disseminate dashboards, reports, and tools that enable the team and company to answer safety-related questions independently
Develop safety data flywheel and provide safety research with production insights/data for training and evaluation.
Qualifications
Minimum
5+ years experience in a quantitative role navigating highly ambiguous environments, ideally as a founding data scientist or team lead at a hyper-growth product company or research org
Proven leadership skills, including leading multiple data scientists and cross-functional teams
Expertise in defining and implementing metrics, with a track record of operationalizing new feature and product-level metrics from scratch
Excellent communication skills with demonstrated ability to communicate with product managers, engineers, and executives alike
Strategic insights that extend beyond traditional statistical significance testing.
Preferred
Experience in trust and safety, integrity, anti-abuse, or related fields
Demonstrated prior experience in NLP, large language models, or generative AI
Strong statistical background, including knowledge of sampling, regression, causal analysis, and more