Inspiration
Millions of people fall victim to scams via WhatsApp, SMS, emails and fake websites every day. Traditional fraud detection tools are slow or overly technical, leaving everyday users vulnerable. We wanted to create a fast, reliable and explainable AI system that anyone can use to instantly check messages and links before taking action.
What it does
ProofCheck AI allows users to paste a text message or website URL and get: Risk classification (Low, Medium, High) Explainable analysis highlighting risky phrases and patterns Actionable advice like ignore, block or report
It uses a hybrid approach, combining instant rule-based detection with AI analysis to ensure both speed and intelligence. Even if AI services are slow or unavailable, the system continues to provide reliable results.
How I built it
Frontend: HTML, CSS, JavaScript for a clean responsive interface Backend: PHP REST API handling rule-based detection and AI requests AI Layer: Gemini 2.5 Flash NLP model for pattern recognition and scam classification Architecture: Parallel hybrid processing to minimize latency, smart caching, fallback to rule-based analysis
Performance optimizations: Token limits, low-temperature AI settings, progressive client-side feedback for smooth UX
Challenges I ran into
Balancing speed and accuracy in AI analysis Ensuring explainable outputs for users without overloading them with technical details Handling fallbacks gracefully if AI API is unavailable Designing a demo-ready system suitable for hackathon presentation
Accomplishments that I'm proud of
Built a working MVP with hybrid AI + rules, fully deployable Created explainable and user-friendly results, not just a black-box AI Optimized the system so it feels instantaneous, even with AI processing Produced a demo-ready system with sample scam messages for instant testing
What I learned
How to combine rule-based and AI methods to optimize real-world systems Importance of explainability and user trust in AI-driven applications Practical performance optimization techniques for AI APIs How to structure a hackathon project to demonstrate real-world impact
What's next for ProofCheck AI
Implement multi-language support, starting with local African languages Real-time streaming for sub-second AI analysis Integrate user feedback reporting to continuously improve detection patterns Explore mobile app deployment for wider accessibility
Log in or sign up for Devpost to join the conversation.