Case Study: Federal Employment Agency achieves improved code quality and fewer errors with Capgemini machine learning code analysis

Machine Learning for Code Analysis at the Federal Employment Agency

The Federal Employment Agency (FEA) in Germany faced a challenge in ensuring the quality of its massive, 800,000-line system responsible for distributing €25 billion in unemployment benefits annually. Despite using conventional static code analysis, it was difficult to identify complex errors that didn't violate pre-defined rules, leaving vulnerabilities that could lead to software failures. Capgemini was engaged to address this need for enhanced quality assurance.

Capgemini, collaborating with the FEA's IT systems integrator and the University of Potsdam, developed a machine learning-based static code analysis tool. This solution automatically identified error patterns and new rules from verified source code, allowing the agency to find and correct errors before they could cause damage. The results included optimized quality assurance, the avoidance of hotfixes, increased system security, and the prevention of software failures, significantly enhancing the system's efficiency and reliability for the Federal Employment Agency.

Federal Employment Agency

Thomas Paal

Business Unit Leader