Fraud Data Scientist

About the role

As a Fraud Data Scientist, you will be a core technical contributor within Billie's Decision Science group. You will design and build robust, scalable machine learning solutions that prevent fraud, with a direct and measurable impact on Billie's bottom line.

What you'll do

• Design and ship anti-fraud models, taking ownership of project priorities and delivering production-ready solutions.
• Model debtor behavioral patterns, identify risk factors, and optimize the logic of Billie's real-time decision engine using quantitative analysis, data mining, and advanced ML.
• Balance precision and recall under severe class imbalance, explicitly weighing the cost of false positives (customer friction) against missed fraud (financial loss).
• Monitor deployed models for drift and adversarial adaptation, and retrain or recalibrate as fraud patterns shift.
• Collaborate with data and software engineers, analysts, and product managers to improve decision logic, integrate new data sources, and extend system functionality.
• Own the deployment and operationalization of ML services within real-time latency constraints, working with Engineering on infrastructure requirements such as containerization and event-driven architectures.
• Turn technical findings into clear, actionable recommendations through effective data storytelling for both technical and non-technical stakeholders.

What you'll need

• 3-5+ years in a quantitative or machine learning role, ideally in fintech or another high-transaction environment. Direct experience in fraud prevention or risk modeling is strongly preferred.
• Proven advanced proficiency in Python (e.g. pandas, scikit-learn, xgboost) and SQL (Snowflake, Postgres, or MySQL).
• Deep expertise in classification models (classical and deep learning), anomaly detection, and graph-based methods (e.g., graph neural networks, entity-link analysis).
• Hands-on experience productionizing ML services, with a strong grasp of modern MLOps concepts such as containerization (Docker/Kubernetes) and event-driven architectures.
• Proven ability to manage stakeholders across technical and non-technical functions, aligning technical roadmaps with business priorities.
• Sharp problem-solving skills, with the ability to translate complex business challenges into clean, efficient, and scalable technical requirements.

Nice to have

• Experience with ML orchestration frameworks such as Metaflow, Apache Flink, or similar MLOps tooling.
• Experience implementing LLM-based workflows (e.g., agentic pipelines, retrieval-augmented generation, or LLM-assisted feature extraction), particularly applied to fraud detection or risk signals.

What they offer

• Virtual Shares Incentive Program
• Flexible work hours and hybrid working approach (up to 3 days per week from home)
• 30 days vacation per year, sabbatical opportunities, and extra child sickness leave
• Yearly development budget and free German group classes
• Discounted access to Berlin Public Transport, Deutschland-Ticket, or JobRad
• Company and team events, interest groups, and multicultural team with 40+ nationalities