Fernando Failla

Built two custom retail Probability of Default models matching off-the-shelf performance, redirected strategy to shelf delivery to avoid dedicated deployment costs, and developed off-the-shelf behavior-based models that delivered a 3-point KS lift on average while contributing to company-wide R package automation and AutoML improvements.

Led automation of a backtest validation pipeline and data validation systems processing 161 requests in 1.5 hours, built a data quality/governance pipeline deployed to 13 products, and developed feature-store tooling that reduced feature volume 70% while retaining 99% predictive performance; delivered an anti-fraud model that raised approval rate from 77% to 85% with R$450,000 estimated incrementa

Developed propensity and unsupervised clustering models for insurance acquisition and segmentation, implemented AutoML pipelines that reduced development cycles from months to weeks, and maintained a centralized Feature Store to accelerate model development and monitor production performance.

Automated SAS processes and operational tasks to monitor delinquency indicators, prepared monthly closing reports for vehicle products, developed dashboards to identify delinquency drivers and fraud indicators, and reduced recurring manual work via SAS Macros and VBA.

Bachelor's in Actuarial Science with a thesis on parametric survival analysis models for hospital bed management in the Brazilian public health system (SUS); thesis available in the UNIFESP repository.