LaunderGraph: AI-Powered AML Investigation

Money laundering can involve complex transaction patterns that are difficult to identify through manual review alone. Financial institutions and investigators may need to examine large volumes of transactions while distinguishing genuinely suspicious activity from normal financial behavior. This creates challenges in scalability, prioritization, and explainability. LaunderGraph addresses this problem through an AI-powered Anti-Money Laundering investigation platform. The system processes transaction data, constructs relationships between accounts, engineers transaction and graph-based features, and applies machine-learning techniques to identify transactions that may require further investigation. Instead of simply producing a binary prediction, LaunderGraph provides an investigation-oriented risk output. Each flagged transaction includes a risk score, risk level, and contributing reasons so that investigators can understand why the system considered a transaction suspicious. The system also provides an interactive dashboard where users can view flagged transactions, search and filter cases, examine transaction details, and organize investigations. The target users are AML analysts, financial investigators, compliance teams, and organizations that need to review large transaction datasets efficiently. Users interact with the system through the dashboard, where they can explore flagged transactions, inspect account-to-account activity, review risk explanations, and manage investigation information. What makes LaunderGraph distinctive is its combination of transaction analysis, graph-based financial relationships, machine-learning detection, and explainability within one investigation workflow. Rather than treating the model as a black box, the platform exposes meaningful reasons behind its risk signals, helping investigators focus their attention where it is most needed. The system is designed as a modular pipeline: transaction preprocessing, feature engineering, graph construction, ML-based detection, explainability, backend APIs, and an interactive frontend dashboard. This allows the analytical pipeline and investigation interface to work together while remaining extensible for future improvements. LaunderGraph is intended to support investigators rather than replace human judgment. A flagged or high-risk transaction is an investigation signal, not proof of money laundering. The goal is to reduce the effort required to identify and investigate potentially suspicious activity while making the process more transparent and scalable