Sr Data Scientist - US Remote

MoneyGram•New York, NY

6d•$130,000 - $185,000•Remote

About The Position

The role involves developing advanced fraud detection solutions using machine learning and data science techniques. Responsibilities include building models with gradient boosting and exploring deep learning approaches, designing supervised and unsupervised anomaly detection systems, and engineering features from transactional, behavioral, and identity data. The position requires deploying models into real-time production environments with scoring and explainability, conducting champion/challenger experiments, and creating monitoring dashboards for performance, drift detection, and feature stability. The individual will analyze fraud patterns across corridors, customer segments, and transaction types, investigate false positives and negatives, and optimize trade-offs between approval rates and fraud losses. Additional duties include documenting model architecture and performance, supporting data labeling strategies, and communicating insights to both technical and non-technical audiences.

Requirements

4+ years of experience in machine learning and data science
2+ years building production ML models in fraud, risk, payments, or financial services
Proven experience deploying and maintaining models in real-time production systems
Strong proficiency with gradient boosting frameworks (XGBoost, LightGBM, CatBoost)
Solid feature engineering skills—ability to extract signal from transactional and behavioral data
Production ML experience including model serialization, deployment, and performance monitoring
Proficient SQL for working with large datasets (BigQuery, Snowflake, or similar)
Proficiency in Python: pandas, NumPy, scikit-learn, and familiarity with deployment tools
Understanding of model explainability (SHAP values, feature importance, gain importance)
Experience with A/B testing or champion/challenger experimental design
Understanding of fraud detection concepts: false positive/negative trade-offs, precision/recall, threshold optimization
Familiarity with common fraud signals (velocity, device, identity, behavioral)
Ability to translate model outputs into business impact (approval rates, loss rates, customer friction)

Nice To Haves

Experience with payment fraud, account takeover, or identity fraud specifically
Familiarity with identity verification signals (device fingerprinting, phone/email risk)
Experience with decisioning platforms (Oscilar, Datavisor, Actimize, or similar)
Background in anomaly detection or unsupervised learning for emerging fraud patterns

Responsibilities

Build fraud detection models using gradient boosting and experiment with deep learning approaches where appropriate
Build supervised / unsupervised anomaly detection models on labeled and unlabeled data to identify outliers
Engineer features from transaction history, device fingerprints, behavioral signals, and identity data
Deploy models to production with real-time scoring and reason code generation
Design and execute champion/challenger experiments to validate model improvements
Build monitoring dashboards for model performance, drift detection, and feature stability
Analyze fraud patterns by corridor, customer segment, and transaction type to identify modeling opportunities
Investigate false positives and false negatives to drive continuous model improvement
Quantify trade-offs between approval rates and fraud losses at different threshold levels
Document model architecture, feature definitions, and performance characteristics
Support data labeling strategy and quality assurance with operations teams
Communicate insights and recommendations clearly to both technical and non-technical audiences