🚀 FraudShield AI Detecting fraud in milliseconds before it detects you. 💡 Inspiration With the rapid rise of digital payments, financial fraud is increasing at an alarming rate. We wanted to build a real-time AI system that can detect fraudulent transactions instantly and prevent financial losses before they occur. 🛠️ How We Built It

FraudShield AI uses a cloud-native architecture on AWS with real-time data streaming and machine learning. We trained a binary classification model:

𝑦

{1Fraud 0Legitimate y={ 1 0 Fraud Legitimate If: 𝑃(Fraud∣𝑋)>0.8 P(Fraud∣X)>0.8 the transaction is flagged immediately

Built With

python

Submitted to

Amazon Nova AI Hackathon

Created by

My Contribution
I led the design and development of the core fraud detection system. My primary contributions included:
Designing the machine learning pipeline for binary fraud classification
Implementing the model in Python and optimizing prediction thresholds
Integrating AWS services such as SageMaker, Lambda, and Kinesis
Developing the backend logic for real-time transaction processing
Handling feature engineering and data preprocessing
Preparing documentation, testing instructions, and final project submissionI also contributed to debugging, performance optimization, and ensuring the system achieved low-latency fraud detection suitable for real-world deployment.

Madiha Mahvish

Updates

Madiha Mahvish started this project — Feb 14, 2026 09:56 PM EST

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.