-
-
UI
-
-
Inspiration
The inspiration for this project came from a prototype fraud detection engine. The main objective was to create a powerful tool capable of detecting fraudulent transactions on a company's payment cards while also providing an expense management system.
What It Does
After completing the first challenge, I revised my vision for this project and decided to focus on polishing the user experience of Falcon. I integrated several frameworks and tools to make the platform more intuitive and user-friendly.
How We Built It
The system is distributed through Docker and consists of four services: Frontend, Backend, Engine, and Tuner.
The Frontend is built with React, the Backend is written in NestJS, the Engine is developed in Python, and the Tuner uses Flask. Not all services communicate directly with each other. The Tuner is designed to be accessible only to authorized personnel responsible for adjusting the engine's thresholds and configuration.
The Engine uses two separate TCP socket servers: one for receiving CSV files and returning fraud analysis results, and another dedicated to configuration management.
Challenges We Ran Into
The main challenges were related to frontend development. I wanted to provide the best possible user experience, which led to multiple rounds of refactoring and significant design changes before reaching the final result.
Accomplishments That We're Proud Of
The engine appears to detect fraudulent transactions accurately within the provided test data. Although we lacked sufficient data to calculate a reliable F1 score, the results remain impressive.
The tuning system was also a significant achievement, as configuration changes can be applied without restarting the service, ensuring minimal downtime and a smoother workflow.
What We Learned
I learned a great deal about fraud detection techniques, methodologies, and the various approaches algorithms use to weigh factors and generate a final score that determines whether a transaction is legitimate or fraudulent.
What's Next for Falcon
Although Falcon already includes an AI chatbot to assist users in making decisions, the next step is to integrate AI-driven fraud analysis. An intelligent system capable of reviewing suspicious transactions could help reduce false positives and, more importantly, false negatives.
Log in or sign up for Devpost to join the conversation.