open-paws-fundraising

Fundraising intelligence platform for animal advocacy organizations. Shared service for coalition partners.

What it does

• Donor churn prediction — CatBoost (or sklearn fallback) model scoring supporters by churn risk using behavioral signals only (no PII stored in feature table)
• Grant matching — Keyword + optional semantic matching of org profiles against 20+ real animal welfare grant opportunities
• Application drafting — LLM-powered LOI, executive summary, and narrative outline generation
• Application tracking — SQLite pipeline tracker with win rate analytics
• Revenue forecasting — 12-month projection with confidence intervals using exponential smoothing
• Impact reporting — Campaign outcome → donor-facing impact statement generation

Architecture

src/
├── donors/         # Churn prediction + re-engagement
├── grants/         # Matching, drafting, tracking
├── forecasting/    # Revenue time-series
├── impact/         # Outcome → reporting loop
└── api/            # FastAPI (org-authenticated)
dashboard/          # Streamlit intelligence dashboard
docs/               # Grant database documentation

Coalition Use

Each coalition partner org operates in full isolation — no cross-org donor data visibility. The grant database is shared, so all orgs benefit from the same grant discovery and seeding work.

Quick Start

pip install -e ".[all]"
cp .env.example .env  # add your ANTHROPIC_API_KEY

# Run API
uvicorn src.api.server:app --reload

# Run dashboard
streamlit run dashboard/app.py

Privacy

Donor behavioral features contain zero PII. All donor identifiers are pseudonymous donor_id strings assigned by the coalition partner's CRM. This platform never receives or stores names, email addresses, or contact information.

Security

See CLAUDE.md for the three-adversary threat model applied to this codebase.

Open Issues

See GitHub Issues for the current backlog.