Spotting Scams, Saving Funds
AI-Powered Real-Time Call Scanner

Problem

Annually $40 billion dollars are lost to scam calls. More than 50 billion scam calls happen annually, with a 7% success rate.

The graph above shows the amount of money lost on average for each scam call.

Victims

• Elderly
• College Students
• Your Loved Ones

Our Approach

Users check numbers against a scam database, get warnings if flagged, upload call audio for analysis via Hume API and Anthropic Sonnet, view results on AWS Bedrock, and receive scam prevention education. Live calls are monitored with Hume Websocket for real-time scam alerts. Users can upload past mp3 voice calls and get an scam analysis of the call.

Steps

• Target Receipient Phone Number Entry
• Real-Time Call Processing
• Speaker Diarization Using Hume API.
• Scammer Tone Detection
• Timestamp and Sentiment Logging
• Scam Identification
• Audio Transcription
• User Warning Notifications
• Cross-Analysis on Emotional Sentiment Data
• Detailed Explanation of Suspicious Text and Emotional Analysis.

What Makes Us Unique

Our solution employs Hume AI's real-time emotion analysis to detect scam calls by analyzing caller voices for suspicious cues. It proactively alerts users and offers actionable steps to protect their security. After each interaction, Hume AI provides educational summaries on detected scam indicators, empowering users with ongoing knowledge to recognize and prevent future threats effectively.

Unique Features

• Phone Number Lookup
• Emotion Based Analysis
• Real-time Proactive Alerts
• Educational Summary

Tech Stack

Front-End

• Next.js

Back-End

• Flask

Languages

• Python
• Typescript

AI

• Hume AI
  • Rest API (Processing audio files)
  • Websocket (Analyzing live audio data)
• Claude AI (Anthopic)
• groq

Cloud Services

• AWS S3