Case study

FraudIA Claims · ML Fraud Detection

Functional prototype designed for the Aseguradora del Sur challenge to identify, evaluate, and classify claim anomalies with comprehensive explainability.

Python
scikit-learn
React
FastAPI
Convex

View code Visit

fraudia-claims01 / 04

Overview

Integrates a supervised Random Forest model, heuristic business rules, and a web dashboard for adjuster reviews without automated claims rejections.

The challenge

Detect complex fraud patterns across large claim volumes while minimizing false positives that affect legitimate clients.

The solution

Hybrid machine learning architecture with per-claim explainability, FastAPI backend, Convex reactive storage, and a modern React and Tailwind CSS UI.

Results

Efficient prioritization of high-risk claims, natural language assistant for claim auditing, and drastic reduction in manual review turnaround.

Project gallery

Final project images will be added soon.

Main preview of the project. — Main view02

Image coming soon

Interface detail03

Image coming soon

Project flow04

Back to projects