Staff Data Scientist (Safety & Risk)

About the Team

Uber is looking for an experienced and motivated Scientist to join our Global Safety & Risk Team. In this role, you will significantly contribute to the physical safety and security of millions of Uber users globally. You will apply your knowledge in data analysis, machine learning, and statistical modeling to generate insights and develop new algorithms that identify and prevent safety incidents before they occur.

This position offers the chance to innovate our technology stack by utilizing the latest breakthroughs in Predictive Modeling, Causal Inference, and Real-time Risk Systems to create intricate autonomous safety frameworks throughout the Uber ecosystem.

About the Role

• Conduct thorough analyses of large, highly imbalanced datasets to identify trends, patterns, and opportunities for improving safety incident detection.
• Design, implement, and optimize binary classification models and algorithms to predict the probability of high-severity incidents.
• Generate actionable insights from risk data and communicate findings to stakeholders, balancing safety interventions with marketplace growth.
• Work as a thought expert for your cross-functional partners (Product, Ops, and Engineering), pushing the boundaries of how Uber defines and mitigates risk.
• Design complex experiments (Diff-in-Diff, Synthetic Controls) and interpret results to draw impactful conclusions in a marketplace environment where classical A/B testing is often not feasible.
• Define how your cross-functional team measures success by developing Safety & Risk metrics (e.g., Recall, Precision-Recall AUC, Probability Calibration) in partnership with global stakeholders.
• Stay current with the latest advancements in Supervised Learning, Causal Inference, and Anomaly Detection.

Technical Skills

Required

• Senior and/or Staff seniority working as a Data Scientist, Applied Scientist, or Machine Learning Engineer.
• Experience building and deploying Binary Classification systems in production for large-scale, high-stakes applications (e.g., Fraud, Safety, or Risk).
• Deep experience in Experimental Design beyond classical A/B testing, including Quasi-experiments (Diff-in-Diff, Synthetic Control) or Marketplace experiments (Switchbacks).
• Expertise in handling Extreme Class Imbalance and optimizing models for rare event detection.
• Experience using Python and SQL to work with massive, high-dimensional data sets at scale.
• Solid foundation in Statistical Methodologies such as probability calibration, sampling, and hypothesis testing.

Preferred

• Experience with Geospatial data analysis (e.g., H3, S2 geometry).
• Background in Real-time Inference systems (working with streaming data like Kafka or Flink).
• Knowledge of Causal Inference to measure the incremental impact of safety interventions.

Please note: this hybrid position is based in São Paulo or Rio de Janeiro, Brazil - welcoming both local professionals and those open to relocating to those cities.