Inspiration
In India, thousands of retail investors and everyday citizens lose money daily to fraudulent investment schemes, WhatsApp scam links, and fake trading groups. ScamShield Bharat was created to give everyone an instant, AI-powered tool to analyze suspicious messages and link parameters before falling victim to fraud.
What it does
ScamShield Bharat allows users to paste suspicious text, links, or investment promises. Using advanced AI models provided by NVIDIA, the system analyzes the message structure, identifies red flags, calculates a risk level, and gives users clear, safe verification actions in simple language.
How we built it
- Frontend & UI: Built with React, Next.js, and Tailwind CSS, hosted seamlessly on Vercel.
- AI Core: Integrated NVIDIA's open-source
nvidia/nemotron-4-340b-instructmodel via the NVIDIA NIM API to run multi-lingual fraud detection and risk assessment. - Backend API: Created serverless API routes to securely handle user input and process responses safely.
Challenges we faced
Migrating from previous model providers to NVIDIA NIM endpoints while ensuring low-latency responses for real-time scam scanning and setting up zero-trust fallback heuristics.
What we learned
We learned how to leverage NVIDIA open-source foundation models for specialized safety and fraud detection tasks, as well as optimizing prompt structures for reliable multi-lingual risk analysis.
What's next for ScamShield Bharat
We plan to add automated WhatsApp bot integration so users can forward suspicious messages directly from their phones for instant scam verification.
Log in or sign up for Devpost to join the conversation.