Case Study: ABN AMRO Bank

Case Study: ABN AMRO Bank

At ABN AMRO Bank, I was responsible for building a data enrichment service for transactions sourced from all banks using PSD2 data. The data is sourced from competitors or ABN AMRO itself. We build the service for a massive scale reaching up to a million transactions per hour. Inside the bank, multiple consumers are desiring to use the service like banking, investing, and mortgages.

The team was a new team starting alongside an existing team. I helped set boundaries between the teams to create independent teams that could develop and deliver on their own.

For our new team, there were a lot of unknowns. To solve this problem,  I took a strong leading position to adopt an Agile on delivering value first and reduce scope.  We implemented improvements to the daily standup that allowed us to reduce cycle time to average one day. I collaborated with the Product Owner to set priorities aligned with business value and to quickly discover the domain.  Next to this, I prepared the user stories for sprints that would lead to a minimal viable product.

Our service is build using serverless functions written in Java and Python. We were a full DevOps team. I trained my colleagues on how to build software in Python. I introduced techniques like Test-Driven Development and Pair Programming to reduce cycle time. To minimize coupling and speed up testing, we used Inversion of Control. I introduced new CI/CD standards for Python that are adopted inside the whole bank. I revitalized the Python triangle within my first two months at the bank.

Initially, we build the service in AWS. Our team experienced a slow development of infrastructure. The cause of this was convoluted and a monolith of AWS CloudFormation code. The team did not have much experience in writing Infrastructure as Code(IaC), so I introduced AWS CDK in favor of AWS CloudFormation. AWS CDK allows natural IaC written in Python. AWS CDK helped our developers to develop faster on a higher abstraction level. The code became more modular and deployed was easily automated using CI/CD Pipelines.

Show Comments