Senior Data Scientist - Fraud Model Validation
Pubblicato il 28-09-2026 - Klarna in Bardi
In this second-line role, you validate fraud models used to protect payments, logins, and identity at Klarna. You’ll reproduce results, stress-test methodologies, and ensure governance and production readiness across the full model lifecycle. You work closely with first-line teams to surface risks and maintain model trust at scale. You’ll shape validation tooling and agentic AI workflows to keep validation pace with rapid development. This is a mission-driven opportunity to strengthen fraud defenses at a large, data-driven fintech. competitive compensation hybrid/onsite work (2–3 days in office) diverse, inclusive culture opportunity to work with cutting-edge AI impactful role in safeguarding payments career growth and cross-team collaboration Assess model performance using fraud-specific metrics and balance business trade-offs Review large transaction datasets and feature pipelines for representativeness and leakage Evaluate drift detection, retraining strategies, and production monitoring Assess CI/CD and deployment controls (Docker, Jenkins, AWS) for model environments Evaluate governance documentation, explainability,
and regulatory compliance Validate emerging techniques (graph networks, anomaly detection, GenAI-based systems) and document risks Communicate validation outcomes and risks to data scientists, ML engineers, and stakeholders 3+ years hands-on fraud modeling Fluency in Python and SQL; experience with PySpark or Spark Experience with tree-based models (LightGBM), anomaly detection, graph/network models Experience across ML lifecycle from feature engineering to deployment and monitoring Ability to explain complex models and communicate to non-technical stakeholders Knowledge of model risk governance, bias, fairness, and privacy considerations Experience building or validating agentic AI workflows Premio: advanced degree in quantitative field; domain experience in BNPL or payment products Mentor or lead validation discussions is a plus ability to challenge approaches constructively detail-oriented with risk awareness LightGBM anomaly detection
#J-18808-Ljbffr