Description
At OpenAI, our User Safety & Risk Operations (USRO) team helps protect our products and users from abuse, fraud, safety risks, and other forms of misuse.
As part of the Automation & Analytics function within RIS, you'll partner with teams across USRO to support their data, analysis, measurement, and automation needs.
In this role, you'll:
- Work with teams across USRO to define the questions they need answered and identify work that could benefit from automation.
- Use SQL and Python to investigate operational problems and turn findings into recommendations teams can act on.
- Build and maintain datasets and dashboards that teams can use for recurring analysis.
- Define metrics to assess the quality and impact of USRO's work, and measure whether changes produce the expected improvements.
- Build scripts and API integrations to automate recurring tasks, using LLMs where they improve the workflow.
- Evaluate automated workflows against expert-reviewed examples and existing processes to understand errors and measure their practical value.
- Hand off workflows with clear ownership, maintenance instructions, and a feedback loop for reviewers.
You might thrive in this role if you:
- Have strong SQL skills and can use Python to analyze data and build tools that solve practical problems.
- Have worked with messy (operational) data and can assess its quality and explain how gaps affect your findings.
- Can turn a loosely defined request into a clear problem and choose a practical way to solve it.
- Have built and maintained datasets, dashboards, scripts, or API integrations that other people use in their work.
- Know how to define useful metrics and measure whether changes to a process or tool improve outcomes.
- Can explain the trade-offs involved in workflow automation, including where human judgment is needed.
- Can work with domain experts to define clear assessment criteria and work through disagreements about ambiguous cases.
- Have experience designing evaluations or quality checks and using the results to decide when to launch a tool and how to improve it.
For more fresh new jobs, check out yubhub.co.