Data Scientist

We are looking for a Data Scientist who has owned models end to end: from framing the problem and defining the target variable, through feature engineering and validation, to deployment, monitoring and recalibration. This is not an assistant or analyst-support role. You will be accountable for the performance and stability of the models you ship, and you will be expected to speak fluently about the metrics that prove they work.

You will work on high-impact problems such as propensity, churn, default prediction, underwriting, fraud and forecasting, partnering with business, risk and engineering teams to turn model output into decisions.

What You Will Do

• Build, validate and deploy supervised models (classification and regression) for problems such as propensity modelling, churn prediction, default prediction, credit risk, underwriting, cross-sell and next best product
• Develop credit risk scorecards using WOE/IV-based variable selection and logistic regression, alongside gradient boosting approaches (XGBoost, random forest)
• Design and engineer features from large transactional and behavioural datasets
• Run rigorous validation: cross-validation, out-of-time validation, rank ordering and decile analysis, and calibration checks.
• Monitor models in production using PSI, KS, Gini and AUC, and lead recalibration and model stamping/validation cycles.
• Handle class imbalance using appropriate techniques (e.g., SMOTE, resampling, cost-sensitive learning).
• Build forecasting and anomaly/fraud detection solutions where required (time series, recommendation systems).
• Explain model decisions to stakeholders using SHAP and feature importance, and translate results into business impact.
• Design and analyse A/B tests and experiments to measure model lift.

What You Bring

Must Have

• Hands-on experience in classification, regression, logistic regression, random forest, gradient boosting and XGBoost.
• Strong feature engineering, cross-validation and hyperparameter tuning skills.
• Proficiency in Python (scikit-learn, pandas, NumPy) and SQL.
• Demonstrated model ownership: you can name the target variable, the problem type, and the evaluation metrics (e.g., AUC/ROC AUC, Gini, KS, precision at k, capture rate) for models you have built.

Strong Advantages

• Credit risk scorecard development, WOE/IV, model validation and stamping.
• Experience with PSI monitoring, calibration and out-of-time validation.
• Experience in BFSI, lending, insurance, fintech, telecom, retail or e-commerce.
• Time series forecasting (ARIMA, SARIMA, Prophet) and demand forecasting.
• Anomaly detection, fraud detection, recommendation systems or collaborative filtering.