FiscAI

Inspiration

Many small and informal businesses in Latin America struggle to understand their financial situation or access proper tools for growth. They often lack digital records, financial literacy, or formal structures that would help them secure funding or plan effectively. We wanted to build something that could guide these entrepreneurs — not just by tracking their numbers, but by helping them predict and improve their growth potential. That’s how FiscAI was born.

What it does

FiscAI is a mobile app that helps business owners analyze their finances and estimate their growth potential using AI. By entering key financial metrics like income, expenses, profit margin, and digitalization level, users get a growth score (from red to green) and insights on how to formalize or improve their business. It acts as a financial assistant, offering personalized recommendations, tracking business health, and helping informal businesses move toward formalization and access to credit.

How we built it

Frontend: React Native + Expo (for iOS and Android) Backend: Supabase (authentication, database, and API endpoints) Machine Learning Model: Random Forest Regressor built in Python (scikit-learn), trained on financial and operational data to predict a business’s growth potential MCP (Model Control Point): Implemented with FastMCP running on AWS Lambda for fast, scalable model inference. This allows the app to send business data, trigger the MCP endpoint, and instantly receive a growth score prediction. Integration: The app sends user data (monthly income, expenses, profit margin, digitalization score, etc.) to Supabase, which then calls the Lambda-based MCP endpoint for prediction. Visualization: React Native Chart showing a red–yellow–green gradient ring to visualize growth potential.

Challenges we ran into

Managing real-time updates between Supabase and the app Integrating the Python ML model predictions into the mobile flow Designing a clean and intuitive UI that small business owners could easily understand Keeping the model interpretable — showing what factors most affect a business’s growth potential

Accomplishments that we're proud of

Built a fully functional end-to-end predictive system, connecting our mobile app to a trained machine learning model through a scalable backend. Implemented a Model Control Point (MCP) using FastMCP deployed on AWS Lambda, enabling real-time and cost-efficient inference directly from the app. This allowed us to serve predictions seamlessly without needing a full backend server or manual deployment pipeline. Designed a clean interface suitable for businesses, emphasizing clarity and accessibility. Successfully integrated Supabase authentication, database management, and secure communication with the MCP-based ML API. Created a financial growth indicator that visually communicates a business’s potential through an intuitive graph, making financial health instantly understandable.

What we learned

How to integrate ML models into React Native apps through API endpoints and cloud functions. How to deploy and manage a Model Control Point (MCP) using FastMCP on AWS Lambda, allowing dynamic and serverless inference that keeps the system lightweight and scalable. The importance of clean data and proper feature selection when predicting financial growth potential. Better practices for mobile UI/UX when presenting financial insights in a simple, visual way for non-technical users.

Gateway for Lambda https://d8pgui6dhb.execute-api.us-east-2.amazonaws.com MCP Server running https://FiscMCP.fastmcp.app/mcp