FinFlow Optimizer

About the Project

Inspiration

Traditional microfinance institutions struggle to balance financial sustainability with borrower affordability. Inspired by the UN’s Sustainable Development Goal 8 (Decent Work), we created a tool to democratize access to fair credit using ethical AI.

What It Does

Analyzes non-traditional data points to:

• Predict repayment likelihood
• Generate fair interest rates (6-24% APR range)
• Provide explainable AI insights for loan officers

How We Built It

• Backend: Python/FastAPI, XGBoost, SHAP, synthetic data generation
• Frontend: Next.js, TailwindCSS, TypesScript
• Deployment: Render (backend), Vercel (frontend)

Challenges We Ran Into

• Data scarcity (solved with synthetic datasets)
• Model bias mitigation using Fairlearn
• Optimizing prediction latency for real-time use

Accomplishments

• 92% default prediction accuracy
• 40% reduction in algorithmic bias
• $0 cloud cost MVP built in 24 hours

What We Learned

• Mobile usage patterns better predict repayment than income history
• Ethical AI requires intentional tradeoffs between accuracy and fairness

What's Next

• Improving the frontend design
• Partnering with various communities to encourage usage of the tool

Built With

• Python
• FastAPI
• XGBoost
• Next.js
• Tailwind CSS
• Render
• Vercel

Built With

• fastapi
• next.js
• python
• render
• tailwindcss
• vercel
• xgboost