ScamShield AI

🛡️ Inspiration

Every day, millions of people fall victim to online scams and phishing attacks. Elderly people and students are especially vulnerable. I built ScamShield AI to give everyone a simple, instant way to verify suspicious messages — no technical knowledge required.

🤖 What it does

ScamShield AI is an AI-powered scam detection platform that analyzes suspicious content in multiple formats:

• Text Analysis — Paste any suspicious SMS, email, or message
• Image/Screenshot Analysis — Upload a screenshot or PDF of suspicious content
• Instant Results — Risk score (0-100), verdict (Safe/Suspicious/Scam), reasons, and safety advice

🏗️ How I built it

• Frontend: React + Vite with custom dark cyberpunk UI
• Backend: Python + Flask REST API
• AI: Amazon Nova 2 Lite via AWS Bedrock for text analysis
• Vision AI: Groq LLaMA 4 Scout for image and screenshot analysis

😤 Challenges

• AWS Bedrock token quota limits on new accounts were very restrictive
• Implementing multimodal image analysis without external OCR tools
• Building a UI accessible to non-technical users

🏆 Accomplishments

• Fully working multimodal scam detector built in under 3 days
• Drag-and-drop screenshot analysis — no typing needed
• Clean professional UI for non-technical users

📚 What I learned

• Amazon Nova models via AWS Bedrock API
• Multimodal AI combining text and vision models
• React component architecture and routing

🚀 What's next

• Mobile app for instant WhatsApp screenshot scanning
• Browser extension to flag suspicious links
• Multi-language support for regional scam detection
• Voice scam detection using Amazon Nova 2 Sonic

Built With

• 2
• amazon
• amazon-web-services
• bedrock
• flask
• lite
• nova
• python
• react
• vite