About the RoleAs a People Data Scientist focused on AI fairness and bias testing, you will help establish how OpenAI evaluates AI-assisted People systems and high-impact talent processes. You will design and conduct rigorous assessments to identify, measure, and mitigate potential bias across the lifecycle of models, agents, decision-support tools, and automated workflows.
Your work will span the entire employee life-cycle, such as hiring, performance, promotion, employee development, workforce planning, etc. You will evaluate both technical systems and the broader human-AI decision processes in which they operate, examining not only model performance but also data quality, measurement validity, differential outcomes, human oversight, and unintended consequences.
We're looking for an experienced data scientist or applied researcher who can translate complex fairness questions into defensible evaluation strategies, scalable testing infrastructure, and clear recommendations for technical teams and senior leaders.
This role is preferred to be based in San Francisco, CA.
In this role, you will:- Define and lead fairness and bias-testing strategies for AI-assisted People processes, models, agents, and decision-support systems from development through deployment and ongoing monitoring.
- Design rigorous algorithmic audits and validation studies, including adverse-impact analysis, subgroup and intersectional evaluation, error-rate analysis, calibration, measurement invariance, reliability, criterion-related validity, and sensitivity testing.
- Identify the appropriate fairness criteria for each use case, evaluate tradeoffs among competing definitions of fairness, and clearly document the assumptions, limitations, and residual risks of each approach.
- Evaluate end-to-end human-AI decision systems, including model outputs, user behavior, human overrides, escalation pathways, and whether AI assistance changes the quality, consistency, or equity of decisions.
- Develop evaluation approaches for generative and agentic AI, including test-set design, counterfactual testing, behavioral evaluation, human-rating studies, robustness testing, and analysis of disparate performance across populations and contexts.
- Investigate the sources of observed disparities, including data representation, label and measurement bias, proxy variables, model design, decision thresholds, workflow design, and differential adoption or usage.
- Partner with engineering, People Operations, Legal, Privacy, Security, and People Systems teams to recommend and evaluate mitigations such as data improvements, model changes, threshold adjustments, workflow redesign, monitoring controls, and additional human oversight.
- Build scalable fairness-evaluation infrastructure, including reusable datasets, automated validation pipelines, regression tests, monitoring systems, self-service tools, and standardized reporting.
- Establish research and documentation standards for fairness test plans, dataset and model documentation, validation reports, limitations, monitoring plans, and decision records.
- Translate complex findings into concise, decision-ready narratives, helping leaders understand the significance of identified risks, the strength of the evidence, available mitigation options, and remaining uncertainty.
You might thrive in this role if you have:- Deep expertise in algorithmic fairness, bias measurement, responsible AI, psychometrics, applied statistics, or the evaluation of high-impact decision systems.
- Exceptional strength in research design, measurement, experimentation, causal inference, and statistical modeling.
- Hands-on experience applying methods such as subgroup and intersectional analysis, adverse-impact testing, equalized-odds and equal-opportunity analysis, demographic-parity assessment, calibration analysis, counterfactual testing, measurement invariance, reliability analysis, and validation studies.
- Strong judgment about the limitations of fairness metrics, including the ability to determine which measures are appropriate for a particular decision context rather than applying a single universal definition of fairness.
- Experience evaluating machine-learning models, generative AI systems, agents, or human-AI workflows using quantitative and qualitative evidence.
- High proficiency in Python or R and SQL, with experience working across complex, sensitive, and imperfect datasets.
- Experience building reproducible evaluation pipelines, automated testing frameworks, analytical tools, monitoring systems, or governed research workflows.
- Ability to distinguish statistical disparities from their potential causes and to communicate findings without overstating certainty or making unsupported causal or legal conclusions.
- Ability to work effectively with technical, operational, legal, privacy, and executive stakeholders and influence consequential decisions through evidence and sound judgment.
- Deep curiosity, intellectual humility, strong attention to detail, and a commitment to developing AI systems and organizational processes that work well for people across different backgrounds and circumstances.
Preferred Qualifications- Experience conducting fairness assessments, algorithmic audits, model-risk reviews, adverse-impact analyses, or validation studies in employment or another high-impact domain.
- Familiarity with fairness and model-evaluation tools such as Fairlearn, AI Fairness 360, responsible-AI evaluation frameworks, explainability methods, or comparable internal tooling.
- Experience evaluating large language models, generative AI systems, safety classifiers, or agentic workflows, including behavioral testing and human evaluation.
- Experience with employment selection, talent assessment, psychometrics, organizational research, or the validation of hiring, performance, promotion, or workforce decisions.
- Familiarity with responsible-AI frameworks and emerging requirements related to automated employment decision systems, algorithmic auditing, data privacy, and AI governance.
- Experience creating model cards, dataset documentation, fairness scorecards, audit reports, monitoring plans, or other review artifacts for high-impact systems.
- Advanced degree in Quantitative Psychology, Computer Science, Statistics, Economics, Data Science, Behavioral Science, or a related quantitative field; PhD preferred but not required.
