A leading global investment bank wanted to automate the manual risk scoring process of its clients and identify the high, medium, and low risks to comply with financial crime compliance (FCC) regulatory requirements. They wanted to categorize risk based on different criteria such as the origin of the money, spending activities and, location of the transactions. The existing risk assessment method was excel-based and risks were computed manually for every division. Some of the other challenges were low quality data, use of legacy system and non standardized batch framework.
We collected customer data from more than 30 sources such as such as accounts, principals, entities, and mortgages and ran it in batches daily, weekly, and monthly. The processed high-quality data was put into Db2, a relational database software where we scheduled data-pull jobs on a Java architecture and create objects. The data was then sent to a different component called factor calculator jobs which calculated risks by extracting division-specific data. The risk scores were displayed on ElasticSearch through web portals against each of the customer profiles. We also developed a standard batch framework that the team could use in the future for sourcing data in a standardized way.