BBVA

Solution

BBVA evolved its ADA global data and AI platform on AWS to industrialize machine learning development and deployment across the bank. Users start from an ADA catalog template that creates a new project and captures regulatory metadata such as model description, business owner, and model tier. The platform automatically provisions a dedicated GitHub repository with standardized code scaffolding for training, model registration, metrics, and metadata. Amazon SageMaker AI provides the core ML workflow foundation for training, batch inference, monitoring, and reusable pipeline components, while ephemeral pull-request environments let teams validate changes in isolated sandboxes before merge. GitHub Actions manages CI/CD transitions across development and production branches. SageMaker Model Registry enforces lifecycle stages and approval states, while Amazon EventBridge and AWS Lambda automate approval workflows, promotions, and production registration. ADA also notifies BBVA’s risk portal on key lifecycle events to keep governance and audit processes synchronized.

Data flow

A user selects an ADA MLOps template, which captures regulatory metadata and provisions a GitHub repository with standardized scaffolding. Teams build and test training, batch inference, and monitoring pipelines on Amazon SageMaker AI, including ephemeral pull-request sandboxes created by GitHub Actions. Models, metrics, and metadata are registered in SageMaker Model Registry, then GitHub Actions, Amazon EventBridge, and AWS Lambda coordinate approvals, promotion, and production registration while notifying BBVA’s risk platform.