About the job
As a Machine Learning Scorecard Developer at Tyme, you will play a pivotal role in designing and optimizing credit scoring components and calibrations that are critical for our approval and risk management strategies across diverse products and markets. Your focus will be on creating high-quality features that ensure score stability and interpretability while delivering comprehensive Probability of Default (PD) to bad rate calibrations that convert model outputs into actionable risk measures.
Key Responsibilities
• Develop and enhance scoring solutions and supporting artifacts utilized in credit decision-making, including application and behavioral scoring, segmentation, and risk signal generation.
• Lead feature engineering for scoring by creating, testing, and documenting variables derived from bureau, application, transactional, and repayment data, ensuring high standards of stability, interpretability, and data quality.
• Contribute to model development and fine-tuning using cutting-edge machine learning techniques, ensuring outputs are robust and reliable for decision-making.
• Implement industry-leading machine learning practices for credit scoring, including systematic hyperparameter optimization, rigorous validation, and repeatable model selection workflows suitable for production environments.
• Define and maintain feature specifications for production, detailing definitions, transformations, edge-case handling, missing value logic, and consistency checks.
• Generate PD and score calibrations based on observed bad rates, including calibration curves, stability tracking, and recommendations for recalibration.
• Assist in analyzing cut-off and limit strategies using calibrated risk outputs to evaluate trade-offs between approval rates, bad rates, and losses.
• Conduct ongoing monitoring of input drift, feature stability, score distribution shifts, performance by segment and cohort, and the health of data pipelines.
• Collaborate with Engineering and Decisioning teams to operationalize scoring outputs, ensuring reproducibility through versioning, back-testing, and change control.
• Maintain comprehensive documentation suitable for internal review and audit, including feature catalogs, calibration methodologies, monitoring reports, and change logs.