Senior/Lead Data Scientist -Credit Risk Modeling
Pubblicato il 03-08-2026 - Klarna in Vicenza
At Klarna, our credit risk models sit at the heart of how we underwrite and price risk for millions of consumers globally.
We're looking for a Lead Data Scientist to help shape the next generation of consumer-level credit scoring and portfolio valuation models.
What You'll Do
As a Lead Data Scientist within credit risk modeling, you will shape Klarna's next-generation consumer-level credit scoring and portfolio valuation models.
You'll design and maintain real-time PD (Probability of Default) models using statistical and ML approaches, integrating them into frameworks for underwriting and economic return optimization.
You'll develop calibration frameworks, ensure compliance with regulatory and fairness standards, and explore novel methodologies — including LLMs for explainability and feature engineering.
Collaborating with cross-functional teams, you'll translate modeling insights into strategic credit policies and business value, while mentoring junior team members and contributing to Klarna's long-term modeling vision.
Who You Are
5+ years' experience in credit risk modeling for consumer lending, credit cards, or BNPL.
Deep proficiency in PD model development and validation, with strong knowledge of calibration techniques.
Advanced Python and SQL skills;
familiar with XGBoost, scikit-learn, pandas, MLFlow.
Experience with explainability frameworks such as SHAP, LIME, PDP.
Ability to communicate technical concepts clearly and influence cross-functional decisions.
Familiarity with real-time modeling and current trends in ML and credit analytics.
Awesome to have
Hands-on experience using LLMs to extract features from unstructured data (e.G., customer communications, credit applications).
Knowledge of integrating third-party credit bureau data into production models.
Understanding of champion/challenger model frameworks and A/B testing infrastructure.
Exposure to loan-level economic modeling, including cost-of-capital and loss metrics.
#J-*****-Ljbffr