Varsha Shah explains why fraud or compliance risks can emerge across payroll, vendor, tax and other records even when each document passes a separate check. She presents a framework that connects entities across sources.
The proposed architecture combines graph-based correlation, adaptive probabilistic risk scores and normalization across jurisdictions. Shah describes investigator feedbackAn AI feedback loop occurs when AI outputs or their consequences become inputs that influence later model or agent behavior. as a way to improve the system over time.
Shah reports evaluationEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria. on approximately three million financial records and describes claimed precision, recall and operational improvements. These are presented as the speaker’s research results, not independently verified performance or a guarantee of fraud detection.
Watch on YouTube
