🎯 The David Protocol

Eradicating the "Credit Invisible" Penalty through AI-Driven Cash-Flow Underwriting.

Traditional credit bureaus penalize millions of financially responsible individuals simply for lacking a history of debt. These credit invisible consumers, including students and gig workers, are unfairly denied essential loans despite maintaining consistent, positive cash flow, creating a harsh cycle of systemic financial exclusion.

The David Protocol securely ingests read-only bank data via the Plaid API. We leverage Gemini's NLP to accurately categorize messy spending habits instantly. Finally, our deterministic machine learning model evaluates cash-flow stability, generating a fair, fully auditable Financial Resilience Score to safely unlock essential credit for millions of untraditional borrowers.

---

⚙️ Tech Stack

Frontend (User Interface & Dashboard)

• Framework: Next.js (App Router)
• Styling: Tailwind CSS + Tremor (Financial UI Components)
• Integration: react-plaid-link

Backend (API & Data Orchestration)

• Framework: FastAPI (Python)
• Server: Uvicorn
• Database: PostgreSQL (Hosted via Supabase/Neon)
• ORM: SQLModel / SQLAlchemy

Artificial Intelligence & Machine Learning

• NLP Categorization: Google Gemini API (Strict JSON Schema outputs)
• Risk Engine: Scikit-Learn / XGBoost (Deterministic Classification)

External APIs

• Financial Data: Plaid API (Sandbox Environment)

---

🧠 How It Works (The Pipeline)

1. Secure Authentication: Users link their bank accounts via the Plaid Link modal (Zero credentials stored on our servers).
2. Data Ingestion: The FastAPI backend fetches up to 90 days of raw transaction history.
3. AI Cleaning (The Smart Layer): Raw, messy transaction strings (e.g., ACH*T-MOBILE WEB PYMT XXXXX992) are batched and sent to the Gemini API, which categorizes them strictly into Rent, Utilities, Income, or Discretionary.
4. Deterministic Scoring (The Safe Layer): The categorized data is evaluated by our XGBoost model against key features (e.g., Income-to-Rent ratio, utility payment consistency) to generate a Financial Resilience Score (1-100).
5. Real-Time Dashboard: The user's score and categorized spending habits are visualized instantly on the Next.js/Tremor frontend.

---

💻 Local Development Setup

Prerequisites

• Node.js: (v18+)
• Python: (3.10+)
• API Keys: Plaid (Sandbox), Google Gemini
• PostgreSQL: Database URL (Supabase or Neon)

🚀 Getting Started (Streamlined)

We have provided PowerShell scripts to automate the setup and runtime process on Windows.

1. Initial Setup

Run the following command in your terminal (at the project root) to set up virtual environments, install all dependencies, and create environment files:

./setup.ps1

2. Running the Project

To launch both the FastAPI backend and Next.js frontend in separate windows:

./run_dev.ps1

---

The David Protocol — Empowering the Credit Invisible.