Tijori — AI-Powered Personal Finance Assistant

AI-driven platform for automated expense tracking, budgeting, and financial insights

Developed an AI-powered personal finance platform that automates expense tracking, categorization, budgeting, and financial recommendations using NLP and LLM systems. Built intelligent pipelines to extract financial data from user emails, enabling passive and seamless expense tracking.

AI & Machine Learning Systems

• Built email-based expense classification system using custom BERT architecture
• Implemented multi-label classification for transaction categories (food, travel, bills, subscriptions)
• Fine-tuned LLaMA (7B) for finance-specific reasoning and advisory responses
• Developed prompt engineering pipelines for personalized financial insights

Email Processing & Data Extraction

• Automated email ingestion system for financial transaction extraction
• NLP-based entity extraction for amounts, merchants, and metadata
• Data cleaning and normalization pipelines for unstructured email data
• Fallback logic for incomplete or ambiguous transaction records

Personal Finance Assistant

• Budget planning and optimization engine
• Spending behavior analysis and insights generation
• Financial goal tracking and recommendation system
• Conversational AI interface for financial queries
• Memory-based personalization for user financial history

Full-Stack Application Development

• React Native mobile app for personal finance tracking
• React + Tailwind web dashboard for analytics
• Node.js backend for authentication and AI inference APIs
• Real-time sync between mobile and web platforms

Cloud Infrastructure & Deployment

• Deployed services on AWS (EC2, SageMaker)
• Scalable ML training and inference pipelines
• Encrypted financial data storage (at rest & in transit)
• High-volume API architecture for transaction processing

Performance & Optimization

• Optimized real-time expense classification latency
• Reduced model size using MLX-based optimizations
• Implemented caching for frequent financial queries
• Improved accuracy through iterative fine-tuning

Security & Privacy

• Secure handling of sensitive financial and email data
• User authentication and session management
• Encrypted API communication
• Privacy-first architecture with minimal data retention

Analytics & Insights

• Spending trends and category breakdown dashboards
• Predictive expense forecasting models
• Behavioral insights for savings optimization
• Exportable financial reports for users

Testing & Validation

• Evaluated classification accuracy for expense models
• Tested LLM financial reasoning outputs
• End-to-end pipeline testing (email → insight generation)
• Scalability testing under high ingestion load