Data Scientist, Safety
$230k - $325k • Remote • San Francisco • FullTime
Posted 4mo ago
About the job
OpenAI's Safety teams are dedicated to ensuring that our products are safe, trusted, and resilient as frontier AI systems scale globally. We are seeking Data Scientists to build the analytical foundations necessary for the responsible deployment of increasingly capable AI. This high-impact role operates at the intersection of product, safety, policy, and research, tackling complex and ambiguous problems where rigorous analysis directly informs critical decisions. Depending on your background and team alignment, you will work on measuring harmful or abusive behavior, detecting fraud and manipulation, evaluating and improving safety systems, designing experiments to understand product impacts, building prevalence estimators and monitoring systems, diagnosing gaps in safety and integrity, and translating risks into measurable problems and evidence-based recommendations. You will partner with Product, Engineering, Policy, Research, and Operations teams to improve safety outcomes and build analytical systems in rapidly evolving domains.
Responsibilities
- Measure harmful or abusive behavior across OpenAI’s products
- Detect fraud, manipulation, and coordinated misuse
- Evaluate and improve safety classifiers, rules systems, mitigation systems, and human review workflows
- Design experiments and causal analyses to understand product, policy, and mitigation impacts
- Build prevalence estimators, dashboards, monitoring systems, and executive decision frameworks
- Diagnose gaps in safety and integrity systems using behavioral and product data
- Quantify and navigate false positive / false negative tradeoffs
- Translate ambiguous safety risks into measurable problems and evidence-based recommendations
- Partner with Product, Engineering, Policy, Research, and Operations teams to improve safety outcomes
- Build zero-to-one analytical systems in rapidly evolving domains
Requirements
- Strong statistical reasoning and analytical judgment
- Experience with experimentation, causal inference, or observational analysis
- Strong SQL and Python skills
- Experience working with messy, incomplete, or noisy datasets
- Ability to structure open-ended business or risk problems
- Excellent communication with technical and non-technical stakeholders
- High ownership and comfort operating independently
- Background in Trust & Safety / Integrity, Fraud & abuse, Security analytics, AI/ML model measurement and evaluation, Alignment and AI safety research, or Biosecurity, synthetic biology, infectious diseases, or computational biology is helpful
Benefits
- Mission-critical problems with world-level impact
- Shape how frontier AI systems are deployed safely
- Unusual ownership and visibility
- Solve novel problems where there are no existing playbooks
- Join highly collaborative teams working across OpenAI