LiquidPay

Inspiration

We were inspired by the common struggle people face when managing finances—especially in low-income or underserved communities. Often, there’s a lack of awareness around where money goes, and no real incentive to improve financial habits. We wanted to build something that doesn’t just track spending but actively motivates better decisions. That’s where LiquidPaywas born—an AI-driven platform that rewards users for building smarter financial behavior.

What it does

LiquidPay is a personalized financial behavior platform that: • Classifies transactions as needs or wants using a machine learning model. • Predicts the user’s financial condition for the upcoming week using a Long Short-Term Memory (LSTM) model trained on the last 60 days of spending data. • Generates daily financial goals based on user behavior using insights from the Gemini API. • Rewards users with our custom ERC-20 token—SaveIT (SIT) for completing goals (5 SIT per completed goal). • Provides a dashboard displaying total balance, savings rate, notifications, and recent transactions, all dynamically updated from backend data.

⸻

How we built it • Frontend: Built with React, providing a clean and interactive dashboard. • Backend: Built with Node.js and MongoDB for storing and retrieving user transactions and goal tracking. • Classification model: Used to categorize transactions into needs and wants before storage. • LSTM model: Trained on 60 days of transaction data to forecast financial conditions for the next 7 days. • Gemini API: Used to analyze user behavior and help tailor daily goals. • Blockchain: Deployed the SaveIT (SIT) token on Polygon Amoy Testnet and implemented reward distribution logic for completed goals.

Challenges we ran into • Integrating machine learning models with backend infrastructure in a seamless and scalable way. • Getting the classification model to reliably tag transactions in real-time. • Ensuring timely generation and deletion of goals, while tracking daily performance accurately. • Connecting blockchain components (especially token transfers and balance checks) using Thirdweb and working through compatibility issues. • Creating a dashboard that is both informative and dynamically linked to real backend data.

Accomplishments that we’re proud of • Successfully integrating AI and blockchain into one cohesive financial tool. • Building a real-time reward system using ERC-20 tokens. • Designing a clean user flow from data classification to behavioral rewards. • Making an impact-driven platform that could genuinely improve financial habits for users.

What we learned • How to apply machine learning models (especially LSTM) to real-world time series financial data. • Hands-on blockchain development—from deploying a token to integrating it into a backend reward system. • How to build a robust system for automated goal setting and tracking using behavioral APIs like Gemini. • Importance of seamless frontend-backend communication for a real-time user experience.

What’s next for LiquidPay • Deploy on a real network and integrate with real financial data providers like Plaid or RazorpayX. • Implement a peer challenge system where users can challenge friends to savings goals for bonus rewards. • Add financial nudges based on LSTM predictions to prevent poor spending weeks. • Launch a mobile-first version to improve accessibility, especially for users in lower-income communities. • Integrate middleman-based cash conversion for token rewards, so even users without crypto knowledge can benefit.

Let me know if you want to add visuals, API code links, or tweak the tone!

Built With

• mern
• polygon
• tensorflow