ArthaRakshak is an AI-powered UPI fraud detection and financial intelligence platform designed to address the growing risks associated with digital payments. With the rapid adoption of UPI, transactions have become seamless, but they have also opened the door to increasingly sophisticated fraud. This project was inspired by the need to build a real-time, intelligent, and transparent system that not only detects fraudulent activity but also empowers users to make safer and more informed financial decisions.

The platform goes beyond traditional systems by integrating multiple advanced capabilities into a single solution. It performs real-time fraud detection, assigns machine learning–based risk scores, and provides explainable AI insights to help users understand why a transaction is flagged. In addition, it includes receiver risk analysis to assess the trustworthiness of recipients, a loan eligibility predictor to support financial decisions, and an RBI circular analyzer to interpret regulatory guidelines, making the system both practical and compliance-aware.

To build this solution, we combined data engineering, machine learning, and user interface design into a cohesive pipeline. We processed transaction data, engineered meaningful features, and trained models such as Random Forest and XGBoost. The system emphasizes explainability and interpretability, ensuring that predictions are not just accurate but also understandable. We developed an interactive interface using Gradio and deployed it on Databricks to ensure scalability and accessibility.

Throughout the process, we addressed challenges such as handling imbalanced fraud data, ensuring consistent model and file sharing across environments, simplifying complex outputs into user-friendly insights, and integrating multiple modules into a seamless interface. Overall, ArthaRakshak transforms fraud detection into a comprehensive financial intelligence platform, combining security, transparency, and decision support in one unified system.

Built With

Share this project:

Updates