DERIV
The Problem: P2P crypto trading is plagued by a simple but devastating scam: The Fake Receipt. Fraudsters use tools like Photoshop or Canva to forge payment proofs, tricking traders into releasing funds before receiving actual payment. Manual verification is slow, error-prone, and unsustainable at scale. The Solution: Deriv P2P Sentinel is an autonomous defense system powered by a swarm of specialized AI agents working in harmony to detect fraud instantly. How It Works: Agent Meta (The Detective): Scans the image file for hidden metadata traces left by editing software (like Photoshop, GIMP, or Canva) and verifies device fingerprints. Agent Privacy (The Guardian): Automatically detects and redacts sensitive PII (Personally Identifiable Information) such as phone numbers and names using OCR and Regex, ensuring data privacy compliance before external processing. Agent Vision (The Analyst): Leverages Google Gemini 1.5 Pro to perform cognitive visual analysis. It spots subtle font inconsistencies, misalignment, and pixel-level artifacts that the human eye misses. Key Features: Sci-Fi HUD Interface: A "Glass-Morphism" tactical dashboard that visualizes the AI's thought process in real-time. Risk Scoring Engine: Aggregates findings from all agents to provide a clear Trust Score (0-100) and actionable verdict (Approve/Review/Reject). Vercel Monorepo Deployment: Fully deployed scalable architecture with Next.js frontend and Python FastAPI backend. Tech Stack: Frontend: Next.js 14, Tailwind CSS, Framer Motion Backend: Python, FastAPI AI/ML: Google Gemini 1.5 Pro, Tesseract OCR Deployment: Vercel (Serverless Functions) Deriv P2P Sentinel transforms compliance from a bottleneck into a competitive advantage, ensuring trust in every transaction.
Category tags: