Inspiration
We were inspired by the rise in scam call reports and the emotional and financial damage they inflict, especially on the elderly and vulnerable. We wanted to build a tool that fights back by turning scammers' own time and energy against them. What better way get scammers back than by giving them a taste of their own medicine?
What it does
ScamSquared is a real-time voice agent that calls scammers and wastes their time while collecting and analyzing conversation data. The goal is to protect potential victims, study scam behavior, and raise awareness using real call content.
How We built it
We used Flask for the backend API and routing logic, and React to build a live dashboard interface. Twilio handles calling and recording, while Deepgram provides transcription. LiveKit lets us stream call audio in real time. The voice agent is orchestrated using Cerebras and Cartesia, enabling dynamic, reactive conversations.
All data is stored in MongoDB Atlas, allowing us to later analyze and surface highlights using vector embeddings.

Log in or sign up for Devpost to join the conversation.