VERDIX AI | Devpost

NEWS ANALYSIS
LINK ANALYSIS
VIRALITY DETECTION
IMAGE DETECTION

Inspiration

VERDIX AI was inspired by real incidents where people close to us were misled by fake news, viral rumors, fraudulent links, and edited government notices circulating on social media platforms. Many of these messages looked official and urgent, causing panic, misinformation, and even financial loss.

We realized that while misinformation spreads rapidly, verification tools are either too technical or fragmented. This motivated us to build a single platform that helps everyday users quickly verify content before trusting or sharing it.

What it does

VERDIX AI is a digital trust verification platform that helps users:

Verify news and textual claims
Detect fraudulent or suspicious links
Analyze virality and manipulation patterns 4.Identify tampered or AI-generated images, especially fake government notices

Instead of giving just a yes/no answer, VERDIX AI provides confidence scores, trust indicators, and source-based verification links, helping users understand 'why' content may be risky or trustworthy.

How we built it

Backend: Built using FastAPI (Python) with a machine learning model trained for fake news detection ML Layer: Combines probability-based classification with heuristic manipulation analysis Verification Engine: Cross-checks claims using trusted online sources and official domains Link Analysis: Detects unsafe URLs, missing HTTPS, and suspicious domains Image Forensics: Integrated a CNN-based image detector hosted on Hugging Face Frontend: Developed using HTML, CSS, and JavaScript with an intuitive mode-based UI Deployment: Backend hosted on Render, frontend deployed via GitHub Pages

Challenges we ran into

Preventing false positives, where fake content could mistakenly appear real
Handling unverified claims when no trusted sources were available
Ensuring smooth frontend–backend communication across hosted platforms
Designing a UI that balances technical depth with simplicity for non-technical users

Accomplishments that we're proud of

Successfully combining ML predictions with real-world verification
Designing a system that only marks content as Real when trusted sources are found
Building a modular verification UI that allows users to analyze content in multiple ways
Creating a solution that focuses on trust, transparency, and explainability