🤝 Top Collaborators
🤓 Latest Submissions
CREDIFY.ai
CREDIFY.ai is an AI-powered scam and fraud warning assistant that helps users identify suspicious digital communications before they lose money or sensitive information. It analyzes SMS/text messages, call transcripts, emails, URLs, and UPI payment requests and provides an explainable risk assessment instead of simply returning a “Scam” or “Safe” label. The system combines multiple detection mechanisms, including machine-learning-based text analysis, rule-based detection, URL analysis, UPI-specific checks, email analysis, call analysis, and threat-intelligence-oriented checks. Its ML pipeline uses TF-IDF word and character features with a calibrated LinearSVC classifier. The model was trained on a curated dataset of 536 scam and safe examples, achieving 98.15% held-out accuracy and 98.51% average 5-fold cross-validation accuracy. These results are limited to the development dataset and are not presented as a guarantee of real-world performance. CREDIFY.ai provides explanations such as risk scores, scam categories, detected red flags, influential tokens, extracted URLs/UPI IDs, and recommended actions. Context-aware rules also help reduce false positives, such as correctly handling genuine OTP messages containing “Do not share this OTP.” The application uses React, Vite, Python, FastAPI, and Pydantic, with responsive UI, dark/light themes, accessibility support, internationalization, and PWA-ready functionality. Privacy is prioritized by avoiding unnecessary personal-data collection and external paid analysis APIs. IBM Bob 2.0 was used during development to understand the existing codebase, improve frontend components and UI/UX, assist with debugging, refine implementation, and accelerate development iterations. Bob was used directly with the project workspace, and Bob session summaries provide evidence of its contribution. CREDIFY.ai aims to act as an intelligent warning layer that helps users pause, understand potential risks, and make more informed decisions before clicking, sharing information, or making payments.
27 Sep 2026
