XAI Audit

Within the KI.SH network I contributed in 2024–2025 to an explainable, AI-assisted solution for the automated verification of budget coverage for the State Audit Office of Schleswig-Holstein. As a student research assistant at IMIS, University of Lübeck, I was responsible for architectural decisions and the ML prototype, including design, implementation and evaluation.

The application translates legal requirements into machine-readable and human-verifiable rule structures. It combines rule-based, graph-based and explainable machine-learning methods so that audit decisions remain traceable and reviewable.

Next steps: testing on additional budget datasets and evaluating the potential to relieve repetitive audit work while keeping human accountability intact. No internal data are displayed on this page.

Technologies: PyTorch, scikit-learn, NumPy, spaCy