Keerthini Prabhakaran

I’m a lead data analyst with a strong foundation in business analytics, statistical modelling, and machine learning, built through hands-on work in the general insurance sector. I focus on turning complex data into clear, decision-ready insights for renewal, pricing, and risk outcomes.

At Royal Sundaram General Insurance, I led end-to-end renewal retention work: from detailed EDA and data integrity checks to feature engineering, encoding, scaling, and model scoring. I designed and validated an XG Boost model for policy renewal likelihood and used SHAP to explain the most critical drivers behind predictions.

I also delivered discount optimization using a monotonic XG Boost approach aligned with underwriting and pricing logic, then operationalized a Python-based optimization model with the SCIP solver. This helped target actions across eligible policies each month, balancing profitability, discount budget, and retention lift.

Beyond retention and optimization, I’ve built and productionized fraud analytics solutions—developing XGBoost fraud scoring models, running Bayesian hyperparameter optimization (Hyperopt), and collaborating with investigation teams to prioritize high-risk “deep-red” claims. I’m passionate about building models that perform in real workflows and communicate value effectively to stakeholders.