Inspriration- Millions of people in rural India struggle with financial literacy because of language barriers, low digital awareness, and lack of trusted information. We wanted to build an AI assistant that speaks their language, explains money in simple words, works offline, and protects users from scams.

About our project- Our system is an AI-powered multilingual financial assistant that: Understands voice and text in multiple Indian languages Teaches financial concepts through simple explanations, stories, and quizzes Provides personalized guidance on saving, loans, and government schemes Works offline in low-connectivity regions Warns users about common scams and frauds Uses a light trust layer to ensure verified financial content

Tech layer Frontend: React (with voice-enabled chat UI) Backend: MERN stack — Node.js, Express, MongoDB AI Layer: Whisper for speech-to-text Indic NLP + translation Llama/Gemma LLM for conversation Vector database for financial knowledge retrieval Offline Mode: ONNX runtime + cached modules Baisic Blockchain: Hash-based verification of financial guidelines Deployment: Cloud-hosted APIs + lightweight mobile-first interface Accurate voice recognition for different accents and dialects

Some of our Challenges Making the system run on low-end devices Ensuring offline functionality while keeping the app lightweight Translating financial terms into simple, culturally relevant language Integrating AI models with the MERN backend smoothly Creating explanations that were accurate yet easy to understand

We learnt- How to integrate speech, NLP, and LLMs into real applications Importance of user-centered design for rural communities Making AI efficient for low-resource devices Structuring a full MERN pipeline for an AI product Building trust mechanisms using verified content How multilingual AI can break huge accessibility barriers

Share this project:

Updates

Submission history