Project Main Image Full-Stack AI Application

ExpenAI - Smart Expense Tracker

A sophisticated web application that leverages OCR and a large language model to automate expense tracking and provide personalized financial insights.

Business Challenge & Mission

Manual expense tracking is tedious, time-consuming, and prone to error, making it difficult for individuals to manage their personal finances effectively. Generic budgeting apps often lack the intelligence to provide truly personalized and actionable advice.

Mission: To create an intelligent, automated expense tracking solution that simplifies data entry through OCR and empowers users with personalized financial guidance from an AI-powered chatbot, making financial wellness accessible to everyone.

Architecture & Project Flow

1. Frontend SPA (User Interface)

A responsive single-page application built with vanilla JavaScript and Tailwind CSS. It provides an interactive dashboard with Chart.js visualizations, user authentication forms, and an intuitive interface for manual expense entry, receipt uploads, and chatbot interaction.

2. Backend API (FastAPI)

A robust and scalable backend built with Python and FastAPI serves as the application's core. It handles user authentication (JWT), provides CRUD APIs for all data, and orchestrates communication between the frontend, database, and AI services.

3. AI OCR Pipeline (Gemini & LangChain)

When a user uploads a receipt image or PDF, the FastAPI backend sends it to a dedicated OCR pipeline. This pipeline uses the Gemini 1.5 Flash model via LangChain to accurately extract and structure key information like expense name, amount, category, and date, returning a clean JSON object.

4. AI Financial Chatbot (Gemini & LangChain)

The chatbot endpoint uses Gemini to provide contextual financial advice. It receives a system prompt, the user's query, and a JSON context containing the user's profile and recent expenses. This allows the LLM to generate personalized, data-driven recommendations.

Key Features

  • Automated OCR Receipt Scanning
  • AI-Powered Financial Assistant Chatbot
  • Interactive Data Visualization Dashboard
  • Secure User Authentication (JWT)
  • Personalized User Profiles & Goals

Technology Stack

Python
FastAPI
Google Gemini
JavaScript
Tailwind CSS
SQLAlchemy

Project Impact

  • Successfully integrates multiple AI services into a cohesive, user-friendly application.
  • Dramatically reduces manual data entry time and friction for the user via OCR.
  • Transforms the app from a simple tracker into a personalized financial co-pilot with its data-driven chatbot.
View Live Application View Project Code