Development of FRIX Fintech Risk Intelligence Platform
Eira Gozum
FRIX — Financial Risk Intelligence Platform
FRIX is an AI-powered fintech risk intelligence platform for real-time fraud detection, transaction risk scoring, mule-risk analysis, and model-driven financial decision support.
The platform currently includes a machine-learning-backed FastAPI service and a premium React dashboard that are integrated through a modular API layer.
Current Project Status
FRIX currently supports:
ML-backed fraud prediction using a trained Random Forest model
Real-time transaction scoring through FastAPI
React dashboard with a premium fintech UI
Frontend-to-backend integration for fraud prediction
Risk level, fraud probability, risk score, model name, and reason-code display
FastAPI test suite with pytest
GitHub Actions CI for backend tests
Dockerized FastAPI service
Environment-based frontend API configuration
Architecture Overview
Repository Structure
Backend — FastAPI Fraud Detection Service
The backend exposes fraud-risk APIs and owns the fraud intelligence logic.
Main endpoints
Run backend locally
From the project root:
Backend runs at:
API docs:
Health check:
Frontend — React Risk Dashboard
The frontend is a React + Vite dashboard that provides the user-facing FRIX console.
Run frontend locally
From the project root:
Frontend runs at:
Frontend Environment Setup
Create a local
.env file inside frontend-react/:An example file is provided:
The local
.env file is ignored by Git.Fraud Prediction Flow
The Fraud Prediction page calls the FastAPI backend through the frontend service layer.
Example request:
Example response:
Testing
Backend tests
Expected result:
Frontend production build
Expected result:
Docker
The FastAPI backend is dockerized.
Build backend image
Run backend container
Stop and remove container
Current ML Model
Current model used by the backend:
Current production-style API supports:
Fraud prediction
Fraud probability
Risk level assignment
Rule-assisted risk score
Reason-code explanation
In CI and Docker test mode, FRIX can use a lightweight mock model through:
This avoids committing large local model artifacts to GitHub.
Planned Enhancements
Future FRIX enhancements include:
Docker Compose for unified frontend + backend startup
API Console page connected to live backend endpoints
Model Monitoring page connected to real metrics
Expanded black-box, white-box, and edge-case tests
Sandbox transaction simulator
Kafka-based transaction streaming
Prometheus and Grafana monitoring
Graph-risk pipeline for mule-network detection
Model selector for Random Forest, XGBoost, LightGBM, graph-risk, and rule-assisted modes
Project Goal
FRIX is being built as a production-style fintech AI platform demonstrating:
Applied machine learning for fraud detection
Full-stack frontend-backend integration
API-first product architecture
Risk explainability
Modular system design
CI/CD and Dockerized deployment foundations
Future-ready streaming and graph intelligence architecture
Like this project
Posted Aug 17, 2026
Developed a fintech platform for fraud detection using AI, FastAPI, and React.
Likes
0
Views
1
