Empowering Financial Literacy Through AI-Driven Multilingual Education
- Overview
- The Problem
- Features
- System Architecture
- Tech Stack
- Installation
- Usage
- Project Structure
- Performance Metrics
- Results and Outcomes
- API Documentation
- Future Scope
- Team
- License
FinShiksha is an AI-powered financial literacy platform designed to bridge the knowledge gap in personal finance across India. Using advanced Retrieval-Augmented Generation (RAG) technology and Groq's Llama 3.3 model, FinShiksha delivers accurate, contextual financial guidance in multiple Indian languages.
Financial literacy is crucial for informed money decisions, yet many Indians lack access to reliable financial education. Language barriers and technological gaps widen this problem further. FinShiksha addresses these challenges by providing personalized, multilingual, AI-driven financial education accessible to everyone.
- Only 27% of Indians are financially literate (FINDEX Survey, 2023)
- Urban Indian households average 2.8 lakh INR in debt (RBI, 2023)
- 66% of Americans fail basic financial literacy tests (FINRA, 2024)
- Average US household debt: $101,915 (Experian, 2023)
Many people lack basic financial knowledge, causing poor money decisions. This leads to financial instability, mounting debt, and limited life opportunities.
- Limited Access: Reliable financial education is not easily accessible.
- Language Barriers: Most resources available only in English.
- Complex Jargon: Financial concepts explained using technical terminology.
- Lack of Personalization: Generic advice that does not address individual needs.
- Scam Vulnerability: Insufficient guidance on identifying financial fraud.
| Region | Financial Literacy Rate | Average Household Debt | Consequences |
|---|---|---|---|
| India | 27% | 2.8 lakh INR | Poor savings, limited investment knowledge |
| USA | 34% (fail basic tests) | $101,915 | Financial stress, limited opportunities |
- Advanced RAG System: Combines factual data retrieval with AI-generated responses for accurate, context-grounded answers.
- Groq's Llama 3.3 Integration: Fast, reliable responses powered by state-of-the-art language models.
- Smart Query Understanding: Natural language processing for intuitive interactions.
- Context-Aware Responses: Maintains conversation history for coherent multi-turn dialogues.
- 3 Languages: English, Hindi, and Kannada support.
- Text and Voice: Both text-based chat and voice playback using gTTS (Google Text-to-Speech).
- Culturally Relevant: Content adapted for Indian financial ecosystem.
- Language Auto-Detection: Automatic language identification from user input.
Verified Sources: Curated data from trusted authorities:
- Reserve Bank of India (RBI)
- Securities and Exchange Board of India (SEBI)
- Income Tax Department
- Insurance Regulatory and Development Authority of India (IRDAI)
- Karnataka Government portals
Topics Covered:
- Banking and Savings Accounts
- Taxation and ITR Filing
- Investments (Stocks, Mutual Funds, Fixed Deposits)
- Insurance and Risk Management
- Scam Awareness and Fraud Prevention
- Government Schemes and Benefits
- Personal Finance Management
- Credit Cards and Loans
- Digital Payment Security
- Live Market Data: Real-time stock market updates and financial news.
- Interactive Visualizations: Charts and badges for user insights.
- Source Citations: Transparent references for all information provided.
- News Aggregation: Latest financial news from verified sources.
- Responsive UI: Clean, minimal interface built with Next.js and React.
- Fast Performance: Asynchronous FastAPI backend for efficient processing.
- Accessible Design: Inclusive interface supporting diverse user demographics.
- Cross-Platform: Works seamlessly on desktop, tablet, and mobile devices.
- Next.js + React: Server-side rendering and responsive UI.
- Multilingual Interface: Language selection and dynamic content.
- Voice Integration: Speech input and audio playback.
- Real-Time Updates: Live market data and news feeds.
- FastAPI Server: Asynchronous request handling.
- RAG Engine: Keyword-based retrieval system.
- Query Handler: Request validation and routing.
- TTS Module: Text-to-speech conversion for multiple languages.
- Knowledge Base: JSON-based multilingual financial data.
- Groq Llama 3.3: Cloud-hosted AI model.
- Market Data APIs: Real-time financial information.
- Government Portals: Verified source integration.
User Query → FastAPI Backend → RAG Retrieval → Knowledge Base Search
↓
Context Extraction → Groq Llama 3.3 API → Response Generation
↓
TTS Processing (if needed) → Formatted Response → User Interface
-
Curated Data Collection: Gathered verified financial data from trusted sources (RBI, SEBI, IT Dept., Govt. portals) with multilingual structuring and official references.
-
Retrieval-Augmented Generation (RAG): Combined factual data retrieval with AI generation for accurate, transparent, and context-grounded answers.
-
Hosted AI Integration: Utilized Groq's Llama 3.3 model for fast, reliable responses without local training overhead.
-
Multilingual and Voice Support: Supported English, Hindi, and Kannada with text and speech (via gTTS) for inclusive accessibility.
-
User-Centric System Design: Built with Next.js/React frontend and FastAPI backend for smooth interaction, efficient processing, and responsive performance.
Framework and Core
- Python 3.8+
- FastAPI (asynchronous, lightweight web framework)
- Uvicorn (ASGI server)
- Pydantic (data validation)
AI and NLP
- Groq API (Llama 3.3 model access)
- gTTS (Google Text-to-Speech)
- Natural Language Processing libraries
Data Management
- JSON-based knowledge base
- Pandas for data processing
- NumPy for numerical operations
API and Integration
- RESTful API architecture
- CORS middleware
- Environment variable management
Framework and Libraries
- Next.js 14
- React 18
- JavaScript/TypeScript
UI and Styling
- Modern CSS
- Responsive design principles
- Component-based architecture
Features
- State management
- Client-side routing
- API integration
- Real-time updates
- Version Control: Git and GitHub
- Environment Management: Python virtual environment
- Package Management: pip (Python), npm (Node.js)
- Testing: Manual and automated validation
- Deployment: Linux cloud infrastructure
- Environment variables for API keys
- HTTPS encryption
- Input validation and sanitization
- Rate limiting
Before installation, ensure you have:
- Python 3.8 or higher
- Node.js 16+ and npm
- Git
- Groq API Key (obtain from https://console.groq.com)
git clone https://github.com/iceman2100/finlit-ai.git
cd finlit-aitext
python -m venv backend/venv
Step 3: Activate Virtual Environment
On Windows:bash backend\venv\Scripts\activate On macOS/Linux:
bash source backend/venv/bin/activate
bash pip install -r requirements.txt
Create a .env file in the project root directory:
bash touch .env Add the following content to .env:
text GROQ_API_KEY=your_groq_api_key_here To obtain your Groq API key:
Visit https://console.groq.com
Sign up or log in
Navigate to API Keys section
Generate a new API key
Copy and paste into .env file
If you have custom training datasets, place them in the datasets/ folder:
text
├── english/
├── hindi/
└── kannada/
If you want to train custom models:
bash python training/train_model.py Note: Pre-trained models are generated automatically on first run if not present.
bash cd backend python -m uvicorn app.main:app --reload The backend server will start at: http://localhost:8000
Step 9: Run Frontend (Open New Terminal) bash cd frontend npm install npm run dev The frontend will start at: http://localhost:3000
Verification Open your browser and navigate to:
Frontend: http://localhost:3000
Backend API Documentation: http://localhost:8000/docs
Usage Starting the Application Ensure both backend and frontend servers are running.
Open browser at http://localhost:3000.
Select preferred language (English, Hindi, or Kannada).
Start asking financial questions.
Example Queries English Examples
"How do I file my income tax return?"
"What are mutual funds and how do they work?"
"How can I protect myself from UPI scams?"
"What is the difference between savings and current account?"
"How do I start investing in the stock market?"
"What government schemes are available for senior citizens?"
Hindi Examples
"मुझे अपना आयकर रिटर्न कैसे फाइल करना है?"
"म्यूचुअल फंड क्या हैं और वे कैसे काम करते हैं?"
"मैं UPI घोटालों से खुद को कैसे बचा सकता हूं?"
"बचत और चालू खाते में क्या अंतर है?"
Kannada Examples
"ನಾನು ನನ್ನ ಆದಾಯ ತೆರಿಗೆ ರಿಟರ್ನ್ ಅನ್ನು ಹೇಗೆ ಫೈಲ್ ಮಾಡಬೇಕು?"
"ಮ್ಯೂಚುಯಲ್ ಫಂಡ್ಗಳೇನು ಮತ್ತು ಅವು ಹೇಗೆ ಕೆಲಸ ಮಾಡುತ್ತವೆ?"
"UPI ವಂಚನೆಗಳಿಂದ ನನ್ನನ್ನು ನಾನು ಹೇಗೆ ರಕ್ಷಿಸಿಕೊಳ್ಳಬಹುದು?"
Using Voice Features Click the microphone icon in the chat interface.
Speak your question clearly.
Wait for transcription.
Receive response with option for audio playback.
Click speaker icon to hear response in selected language.
finlit-ai/
│
├── backend/ # FastAPI backend server
│ ├── app/
│ │ ├── __init__.py
│ │ ├── main.py # Application entry point
│ │ ├── config.py # Configuration settings
│ │ ├── api/ # API route handlers
│ │ │ ├── __init__.py
│ │ │ ├── chat.py # Chat endpoints
│ │ │ ├── tts.py # Text-to-speech endpoints
│ │ │ ├── market.py # Market data endpoints
│ │ │ └── news.py # News endpoints
│ │ ├── models/ # Data models
│ │ │ ├── __init__.py
│ │ │ ├── query.py # Query models
│ │ │ └── response.py # Response models
│ │ ├── services/ # Business logic
│ │ │ ├── __init__.py
│ │ │ ├── rag_engine.py # RAG implementation
│ │ │ ├── llm_service.py # LLM integration
│ │ │ ├── tts_service.py # TTS service
│ │ │ └── knowledge_base.py # Knowledge retrieval
│ │ └── utils/ # Utility functions
│ │ ├── __init__.py
│ │ ├── validators.py
│ │ └── helpers.py
│ ├── venv/ # Virtual environment (not in repo)
│ ├── requirements.txt # Python dependencies
│ └── .env # Environment variables (not in repo)
│
├── frontend/ # Next.js frontend
│ ├── pages/
│ │ ├── _app.js # App wrapper
│ │ ├── index.js # Home page
│ │ └── chat.js # Chat interface
│ ├── components/
│ │ ├── ChatBox.js # Chat component
│ │ ├── LanguageSelector.js # Language switcher
│ │ ├── VoiceInput.js # Voice input component
│ │ ├── MessageList.js # Message display
│ │ └── Header.js # App header
│ ├── styles/
│ │ ├── globals.css # Global styles
│ │ └── Chat.module.css # Chat styles
│ ├── public/
│ │ ├── favicon.ico
│ │ └── images/
│ ├── package.json # Node dependencies
│ ├── package-lock.json
│ └── next.config.js # Next.js config
│
├── training/ # Model training scripts
│ ├── __init__.py
│ ├── train_model.py # Training script
│ ├── preprocess.py # Data preprocessing
│ └── evaluate.py # Model evaluation
│
├── datasets/ # Training data (not in repo)
│ ├── .gitkeep
│ ├── english/
│ │ ├── banking.json
│ │ ├── taxation.json
│ │ └── investments.json
│ ├── hindi/
│ │ ├── banking.json
│ │ └── taxation.json
│ └── kannada/
│ ├── banking.json
│ └── taxation.json
│
├── models/ # Trained models (not in repo)
│ └── .gitkeep
│
├── docs/ # Documentation
│ ├── API.md # API documentation
│ ├── ARCHITECTURE.md # Architecture details
│ └── DEPLOYMENT.md # Deployment guide
│
├── tests/ # Test files
│ ├── test_api.py
│ ├── test_rag.py
│ └── test_tts.py
│
├── .gitignore # Git ignore rules
├── .env.example # Example environment file
├── README.md # Project documentation
├── LICENSE # MIT License
├── CONTRIBUTING.md # Contribution guidelines
└── requirements.txt # Root dependencies
Performance Metrics Experimental Analysis FinShiksha was evaluated to measure multilingual accuracy, retrieval efficiency, response performance, and real-world user experience. The system was tested on a knowledge base built from verified financial sources such as RBI, SEBI, IT Department, IRDAI, and Karnataka Government portals. Experiments were executed on a FastAPI-based backend with Llama-powered advisory generation, deployed via Linux cloud infrastructure for real-time usage.
| Metric | Score | Description |
|---|---|---|
| Retrieval Precision | 0.89 | Accuracy of relevant document retrieval |
| Retrieval Recall | 0.85 | Coverage of relevant documents retrieved |
| ROUGE-1 Score | 0.77 | Unigram overlap with reference responses |
| ROUGE-L Score | 0.74 | Longest common subsequence match |
| BLEU Score (Hindi) | 0.78 | Translation quality for Hindi responses |
| BLEU Score (Kannada) | 0.72 | Translation quality for Kannada responses |
| Average Response Time | 2.3 sec | Time from query to response delivery |
| User Satisfaction | 4.2 / 5.0 | Average user rating |
Performance Benchmarks Response Time Analysis:
Simple queries (definitions): 1.5-2.0 seconds
Complex queries (multi-step explanations): 2.5-3.5 seconds
Voice queries with TTS: 3.0-4.0 seconds
Accuracy by Category:
Banking queries: 92% accuracy
Taxation queries: 88% accuracy
Investment queries: 85% accuracy
Scam awareness: 90% accuracy
Results and Outcomes Impact Summary Enhanced Financial Literacy
Simplified, multilingual explanations improved users' understanding of complex financial topics. Users reported better comprehension of banking products, tax filing procedures, and investment options.
Empowered Decision-Making
AI-backed, factual insights increased user confidence and financial awareness. Survey respondents indicated higher confidence in making financial decisions after using FinShiksha.
Wider Reach and Accessibility
Engaged diverse communities across languages and skill levels, promoting inclusive financial education. Platform reached users from urban and rural areas, with particular success among non-English speakers.
User Feedback Positive Outcomes:
85% of users reported improved financial knowledge.
78% felt more confident about managing personal finances.
92% appreciated multilingual support.
88% found voice features helpful for accessibility.
Areas of Impact:
Better understanding of government schemes.
Increased awareness of financial scams.
Improved knowledge of investment options.
Enhanced tax filing confidence.
API Documentation Base URL text http://localhost:8000/api Endpoints
- Chat Query POST /api/chat
Send a financial query and receive AI-generated response.
Request Body:
json
{
"query": "What is a mutual fund?",
"language": "english",
"user_id": "optional-user-id"
}
Response:
json
{
"response": "A mutual fund is an investment vehicle...",
"sources": [
{
"title": "SEBI Mutual Fund Guidelines",
"url": "https://www.sebi.gov.in/..."
}
],
"language": "english",
"timestamp": "2026-01-02T19:30:00Z"
}
- Text-to-Speech
POST /api/tts
Convert text response to speech in specified language.
Request Body:
json
{
"text": "म्यूचुअल फंड एक निवेश साधन है",
"language": "hindi"
}
Response:
json
{
"audio_url": "/audio/temp_audio_12345.mp3",
"duration": 5.2,
"language": "hindi"
}
- Market Statistics
GET /api/market/stats
Retrieve current market statistics.
Response:
json
{
"sensex": 72845.38,
"nifty": 21793.15,
"change_percent": 0.45,
"last_updated": "2026-01-02T15:30:00Z"
}
- Financial News
GET /api/news?category=banking&limit=5
Get latest financial news articles.
Response:
json
{
"articles": [
{
"title": "RBI announces new banking guidelines",
"description": "Reserve Bank of India releases...",
"url": "https://www.rbi.org.in/...",
"published_at": "2026-01-02T10:00:00Z",
"source": "RBI Official"
}
],
"count": 5
}
- Knowledge Base Search
GET /api/search?query=tax&language=english
Search knowledge base directly.
Response:
json
{
"results": [
{
"title": "Income Tax Filing Guide",
"content": "Income tax return filing is...",
"category": "taxation",
"language": "english",
"relevance_score": 0.92
}
]
}
Error Responses
All endpoints return standard error responses:
json
{
"error": "Error message description",
"status_code": 400,
"timestamp": "2026-01-02T19:30:00Z"
}
Phase 1: Integration and Expansion Banking API Integration: UPI payment, e-KYC, and bank account linking for personalized advice.
Language Expansion: Support for Tamil, Telugu, Bengali, and Malayalam (15+ Indian languages).
Mobile Applications: Native iOS and Android apps with offline mode support.
Phase 2: Advanced Features Voice-Only Mode: Hands-free navigation and complete voice interaction for visually impaired users.
Offline Processing: Local query processing and cached responses for low-connectivity areas.
Community Learning Spaces: Forums and peer-to-peer learning platforms.
Phase 3: Intelligence Enhancement Personalized Recommendations: ML-based advice for portfolios, savings, and budgeting.
Predictive Analytics: Market trend forecasting and risk assessment tools.
Advanced Security: Biometric authentication and blockchain-based transaction security.
Phase 4: Ecosystem Integration Government Portal Integration: Direct ITR filing, Aadhaar/DigiLocker integration.
Financial Institution Partnerships: Direct investment and insurance purchasing capabilities.
