Operational Efficiency and Cloud Scale – BTG Pactual

Overview
BTG Pactual modernized its fraud stack by migrating from on-premises infrastructure to Feedzai on AWS, strengthening security, improving operational efficiency, and unlocking elastic scale. The platform supports millions of daily transactions across cards, Pix, and in-app payments, requiring millisecond decisioning to keep fraud rates low without sacrificing approval rates or customer experience.

Solution
Feedzai delivered advanced machine learning models and its network predictive fraud score (NPFS), now hosted on Feedzai’s cloud platform built on AWS. Real-time fraud scoring uses Amazon EKS and Aurora to deliver consistent, low-latency performance. Model training and feature processing run on Amazon S3 and EMR, allowing data science teams to iterate frequently without touching the live authorization path. AWS Glue ensures data is clean and governed, while Amazon OpenSearch supports fast investigations and operations dashboards. This architecture keeps real-time decisioning isolated, resilient, and scalable while modernizing the rest of BTG’s fraud lifecycle.

One-line takeaway
By running Feedzai on AWS, BTG Pactual achieved millisecond fraud decisioning at national scale, faster model iteration, and lower operational costs, while maintaining low fraud rates and high approval performance across cards, Pix, and app payments.