🚀 ScamTalk: Hackathon Submission

📌 Inspiration

Seniors lost $3.4 billion to scams in 2023 alone, yet they lack real-time scam alerts and trusted fraud education. Traditional media spreads fear but doesn’t provide solutions. Banks like PNC face financial burdens from fraud-related reimbursements and trust erosion. ScamTalk bridges this gap by delivering real-time scam alerts and personalized fraud education to empower seniors and protect their finances.

🔍 What It Does

ScamTalk has two core components:

Scam Alerts – Real-time updates on scams reported by PNC customers, including scam type, amount involved, and trends. Users receive notifications and can track scam patterns to stay vigilant.
Personalized Scam Education – A React-based onboarding system collects basic financial data (age, expenditure, savings, etc.) to create a user profile. Using a profiling algorithm, we identify scams they are most vulnerable to and use DeepSeek R1 (LLM via Ollama) to generate relevant scam-related queries. These queries are used to fetch curated YouTube Shorts to educate users on scams they are likely to encounter.

🛠 How We Built It

Frontend: Built with React, handling onboarding, scam tracking, and notifications.
Backend: Powered by FastAPI, managing scam alerts and user profiles.
LLM Integration: DeepSeek R1 (via Ollama) generates scam-related queries based on user profiles.
Scam Content Curation: Searches YouTube Shorts using scam-specific keywords to provide relevant educational material.
Data Processing: Scam reports are collected and categorized from PNC’s fraud alerts.

🚧 Challenges We Ran Into

Profiling Algorithm Accuracy – Determining the best way to match users with relevant scam threats.
LLM Query Optimization – Ensuring DeepSeek R1 generates precise scam-related queries for better content discovery.
API Limitations – YouTube’s API doesn’t always surface short-form scam awareness content, requiring manual adjustments.

🏆 Accomplishments That We're Proud Of

✅ Real-time scam alerts system that helps seniors stay ahead of fraudsters.
✅ Dynamic scam profiling using a combination of structured financial data and LLM-driven insights.
✅ Curated scam education system that delivers digestible, video-based scam prevention content tailored to users.
✅ A scalable foundation – Future integrations with PNC’s internal fraud detection and educational materials.

📚 What We Learned

🔹 Seniors are overwhelmed by scam information and need clear, concise alerts.
🔹 Banks suffer major financial losses from scam-related reimbursements—prevention is key.
🔹 Profiling scam vulnerability is complex and requires behavioral + financial insights.
🔹 Using AI for scam prevention is promising, but it needs constant refinement to avoid misinformation.

🚀 What’s Next for ScamTalk

Expand scam sources beyond PNC reports—integrate with government scam databases (FTC, CFPB, IC3).
Improve scam profiling by using real-time banking behavior analysis (e.g., transaction monitoring).
Use PNC’s own fraud education content instead of relying on external videos.
Enhance AI-generated scam queries with trusted fact-checking mechanisms.
Deploy as a real banking security feature, reducing scam losses and increasing customer trust.