EWS-Data Scientist

- Design, build, and document Early Warning System (EWS) / behavioral scorecards (retail and corporate/wholesale) using machine learning techniques such as XGBoost, logistic regression, and random forest.

- Perform end-to-end model development: sample design, target definition (e.g. SMA/NPA flagging), feature selection and hypothesis documentation, hyperparameter tuning (e.g. Optuna), cross-validation (k-fold/OOF), and threshold optimization.

- Conduct independent model validation and technical review of vendor-built or in-house models - including checking implementation code against validation reports, verifying variable weights/coefficients, and identifying documentation or governance gaps.

- Monitor live/production scorecards through Out-of-Time (OOT) validation, tracking discriminatory power (Gini, AUROC), score stability, and population drift; flag and investigate anomalies (e.g. OOT Gini exceeding development Gini).

- Prepare Model Validation Documents, Variable Hypothesis reports, and governance checklists in line with AI/ML Model Risk Management (MRM) frameworks.

- Prepare Model Monitoring reports on a quarterly basis.

- Liaise with bank IT teams to translate model logic into implementation-ready specifications

- Present model findings, enhancements, and validation outcomes to Retail Risk, Collections, and other stakeholders at development and each reporting cycle.

- Experience reviewing or validating models built by external vendors.

- Exposure to logistic regression-based legacy scorecards alongside modern ML approaches.

- FRM certification.

- Design, build, and document Early Warning System (EWS) / behavioral scorecards (retail and corporate/wholesale) using machine learning techniques such as XGBoost, logistic regression, and random forest.

- Perform end-to-end model development: sample design, target definition (e.g. SMA/NPA flagging), feature selection and hypothesis documentation, hyperparameter tuning (e.g. Optuna), cross-validation (k-fold/OOF), and threshold optimization.

- Conduct independent model validation and technical review of vendor-built or in-house models - including checking implementation code against validation reports, verifying variable weights/coefficients, and identifying documentation or governance gaps.

- Monitor live/production scorecards through Out-of-Time (OOT) validation, tracking discriminatory power (Gini, AUROC), score stability, and population drift; flag and investigate anomalies (e.g. OOT Gini exceeding development Gini).

- Prepare Model Validation Documents, Variable Hypothesis reports, and governance checklists in line with AI/ML Model Risk Management (MRM) frameworks.

- Prepare Model Monitoring reports on a quarterly basis.

- Liaise with bank IT teams to translate model logic into implementation-ready specifications

- Present model findings, enhancements, and validation outcomes to Retail Risk, Collections, and other stakeholders at development and each reporting cycle.

- Strong grounding in machine learning fundamentals (classification techniques, cross-validation, hyperparameter tuning, evaluation metrics such as Gini/AUROC/KS).

- Proficiency in Python or SAS for model development and data analysis (Python preferred; pandas, scikit-learn, XGBoost, Optuna or equivalent).

- Hands-on experience in credit risk analytics, specifically scorecard modelling (behavioral/EWS), model validation, and ongoing portfolio monitoring.

- Working knowledge of SQL for data extraction and variable sourcing from banking systems.

- Experience preparing formal model documentation - validation reports, variable hypothesis documents, governance/AI-ML checklists - suitable for regulatory/MRM review.

- Ability to independently audit a models implementation (code) against its stated methodology and flag discrepancies.

- Strong written and verbal communication skills for stakeholder presentations and technical write-ups.

- Experience working directly with bank Risk, Collections, and IT teams in a consulting capacity.