Data Science Manager

What you will do Perform independent end-to-end validation of fraud detection ML models, including conceptual soundness, data integrity, feature engineering, model development, deployment design, and monitoring frameworks. Develop challenger models. Review and challenge first-line fraud model methodologies, assumptions, and implementation choices (e.g., scikit-learn, LightGBM, graph models, anonaly detection techniques, GenAI components). Build and deploy agentic AI tools to support model validation workflows — automating review of model documentation and code, surfacing risks and inconsistencies. Assess model performance using appropriate fraud metrics (e.g., precision/recall, ROC-AUC, PR-AUC, cost-sensitive metrics, fraud rate capture, business impact trade-offs). Evaluate model stability, drift detection, retraining strategies, and production monitoring practices. Independently replicate model results where necessary and conduct challenger analyses to assess model robustness and limitations. Review large-scale transaction datasets and feature pipelines (e.g., >100M transactions, hundreds of features) to assess data representativeness, leakage risks, and bias. Evaluate model governance documentation, explainability approaches, and transparency — including regulatory compliance related to model risk, fairness, and data privacy. Validate new technologies applied in fraud detection, such as Graph Networks, Behavioral Biometrics, Anomaly Detection, and GenAI-based systems. Assess controls around CI/CD pipelines, deployment processes (e.g., Docker, Jenkins), and cloud environments (e.g., AWS SageMaker, S3, Athena, Lambda). Develop and maintain validation frameworks, testing standards, and model performance monitoring tools (e.g., SQL, PySpark, Python-based validation libraries). Collaborate closely with first-line fraud data scientists, ML engineers, product, and business stakeholders to ensure transparent communication of model risks and validation findings. Provide a