Paisa Guru

Paisa_guru
Paisa_guru
Paisa_guru
Paisa_guru
Paisa_guru
Paisa_guru
Paisa_guru

Inspiration

First-time and low-literacy digital payment users—our parents, grandparents, and local shopkeepers—are constantly targeted by sophisticated cyber scams like fake KYC expiry calls, UPI collect request traps, and fake customer care numbers. Traditional awareness campaigns are text-heavy, English-first, and completely passive. People only learn how a scam works after they lose their money. We wanted to flip the script: what if people could safely practice resisting a scam before ever facing a real one?

What it does

Paisa Guru is a voice-first AI coach that trains users to spot and resist digital payment scams through realistic voice simulations.

Scam Drills: Alexa plays a realistic scammer scenario (e.g., urgent bank blocking threat). The user replies by voice, and Amazon Nova judges their response in real-time, giving a clear verdict (SAFE, RISKY, or FELL FOR IT) along with a simple explanation.
Message Checker: Users can check suspicious SMS messages to get an instant risk level and clear red-flag breakdowns.
Family Dashboard: A companion web dashboard where students or tech-savvy family members can set up older relatives, track their drill scores, and monitor weekly progress and weak spots.

How we built it

We designed Paisa Guru around Amazon’s developer ecosystem and serverless architecture:

Voice Layer: Alexa Custom Skill built using the Alexa Skills Kit (ASK) SDK for Node.js.
Backend & Logic: AWS Lambda handling core intent routing and API endpoints.
AI & Intelligence: Amazon Bedrock powered by Amazon Nova models for dynamic response judging and scam scenario generation.
Database: Amazon DynamoDB storing user profiles, drill sessions, and score histories.
Frontend: React-based web app for the family dashboard.

Challenges we ran into

Voice Latency Constraints: Alexa requires responses within strict timeout windows. To keep the experience snappy, we pre-generated a robust bank of scam scenarios and stored them, using Amazon Nova live only for real-time response evaluation.
Prompt Engineering for Clarity: Ensuring Amazon Nova's judgments and feedback tips were translated into simple, easy-to-understand language suitable for low-literacy users.

What we learned

Building a voice-first fintech-edu product highlighted how crucial inclusive design is. Accessibility isn't just about screen readers; it's about making security intuitive for people who didn't grow up navigating digital interfaces.

What's next for Paisa Guru

Expanding regional language support (Hindi, Marathi, and regional dialects).
Piloting with local community centers and senior citizen groups.
Exploring partnerships with banks and telecom operators for broader distribution.