Trajectory

# Inflection - AI-Powered Loan Assessment Platform

## Inspiration

After speaking with local business owners, we discovered that entrepreneurs who get denied loans due to past financial mistakes often stop applying, causing their businesses to stagnate and close. Traditional lenders focus too heavily on credit history rather than current financial health and market viability.

## What it does

**Inflection** uses 4 AI agents to assess loan applications by analyzing:
- **Recent Financial Performance** - Real banking data via Plaid to assess cash flow, savings rate, and income stability
- **Market Viability** - Google Maps/Places API to analyze local competition and market conditions
- **Holistic Assessment** - AI agents synthesize all data to generate fair eligibility scores

**Features:**
- Multi-agent AI assessment (Financial Analyst, Market Researcher, Risk Assessor, Coach)
- Plaid integration for real financial data
- Google Maps/Places for market analysis
- AI coach for personalized guidance if not approved
- Action plans and financial visualizations

## How we built it

**Backend:**
- FastAPI
- Google Gemini 1.5 Pro (LLM)
- LangChain
- Plaid API (sandbox)
- Google Maps/Places API
- SQLite with SQLAlchemy

**Frontend:**
- React + TypeScript
- Vite
- Tailwind CSS
- shadcn/ui
- Framer Motion
- Recharts

**Architecture:**
Custom orchestrator runs Financial Analyst and Market Researcher in parallel, then Risk Assessor makes the final decision, and Coach generates recommendations.

## Challenges we ran into

- Coordinating multiple AI agents efficiently
- Mapping frontend data structures to backend APIs
- Learning Plaid's token exchange flow
- Converting Google Places data into meaningful market scores

## Accomplishments that we're proud of

- Built complete 4-agent AI system working together
- Integrated real Plaid and Google Maps APIs
- Implemented parallel processing for faster assessments
- Created secure encryption for financial tokens
- Built beautiful, intuitive UI with smooth animations
- End-to-end functionality from application to recommendations

## What we learned

- Multi-agent architecture design
- Financial API integration (Plaid)
- Location intelligence (Google Maps/Places)
- LLM prompt engineering
- Secure data handling
- Full-stack React + FastAPI integration

## What's next for Inflection

- Move to production Plaid for real financial data
- Add credit bureau API integration
- Build lender dashboard
- Create mobile app