Dashboard
Chat-Bot
Home

💡 Inspiration

Managing personal finances is often tedious because users must manually fill multiple fields for every expense. We wanted to make expense tracking as simple as chatting with a friend.

With the rise of Large Language Models, we envisioned an application where users simply type messages like:

"Spent ₹250 on groceries yesterday"

and the system automatically understands and records the expense.

That's how SpendBoat was born—an intelligent finance assistant that combines AI, Machine Learning, and modern web technologies to simplify expense tracking.

✨ What it does

SpendBoat enables users to record expenses using natural language through either a web dashboard or WhatsApp.

Example:

Spent ₹250 on groceries yesterday

The application automatically extracts:

💰 Amount
🏷️ Category
📅 Date
📝 Description

using Groq's LLaMA 3.

To ensure high reliability, if the AI model is unavailable or rate-limited, a TF-IDF + Logistic Regression model predicts the expense category, achieving 99.9% request success.

Additional features include:

🤖 AI-powered expense parsing
💬 WhatsApp expense logging
📊 Interactive analytics dashboard
💰 Monthly budget tracking
📧 Budget exceeded email alerts
📈 Spending insights and visualizations

🛠️ How we built it

Frontend

React.js
Tailwind CSS
Recharts

Backend

Node.js
Express.js
MongoDB

AI

Groq API
LLaMA 3

Machine Learning

Python Flask Microservice
TF-IDF Vectorizer
Logistic Regression

Integrations

Twilio WhatsApp API
Nodemailer

⚙️ System Architecture

              User
                │
      Web Dashboard / WhatsApp
                │
                ▼
        Node.js + Express API
                │
                ▼
       Groq (LLaMA 3 Parsing)
          │             │
      Success        Failure
          │             ▼
          │     ML Fallback Service
          │ (TF-IDF + Logistic Regression)
          └─────────────┬──────────────┘
                        ▼
                    MongoDB
                        │
                        ▼
              React Analytics Dashboard

🚧 Challenges we ran into

Parsing inconsistent natural language inputs accurately.
Prompt engineering to generate structured JSON responses.
Handling API failures and rate limits gracefully.
Designing an ML fallback system without affecting user experience.
Integrating Twilio WhatsApp webhooks with the backend.
Building responsive dashboards for analytics.
Maintaining fast response times while ensuring reliability.

🏆 Accomplishments that we're proud of

🚀 Built a complete AI-powered expense tracking platform.
🤖 Successfully integrated Groq LLaMA 3 into a production-ready workflow.
🔄 Achieved 99.9% request reliability using an intelligent ML fallback model.
💬 Enabled expense tracking directly from WhatsApp.
📊 Created an interactive analytics dashboard with meaningful financial insights.
🏗️ Designed a scalable architecture separating AI inference and ML services.

📚 What we learned

Throughout the development of SpendBoat, we gained hands-on experience with:

Large Language Model integration
Prompt engineering
AI-powered application development
MERN Stack architecture
REST API development
Machine Learning deployment
Flask microservices
Twilio webhook integration
Data visualization
Error handling and fallback strategies
Building scalable and resilient systems

🚀 What's next for SpendBoat

We plan to expand SpendBoat into a complete AI financial assistant by adding:

🎙️ Voice-based expense logging
📸 Receipt OCR using computer vision
🤖 Personalized AI spending recommendations
📈 Expense forecasting using predictive analytics
💳 Bank account synchronization
👨‍👩‍👧 Shared family expense management
🌍 Multi-language support
📱 Android & iOS mobile applications
🔐 OAuth authentication
☁️ Cloud-native deployment with CI/CD
📊 Advanced financial reports and insights

❤️ Our Vision

SpendBoat aims to become an intelligent personal finance assistant that helps users effortlessly track, understand, and improve their financial habits using the power of Artificial Intelligence, Machine Learning, and Conversational Interfaces.