Logo Lanfrica

PatrickAttankurugu/trustscore

Domaine:

socioeconomic

Type de record:

software
Créateur:
Pat
Hôte:
AI-powered alternative credit scoring platform for Sub-Saharan Africa. Analyzes mobile money, telecom, utility, and behavioral data to generate credit scores for unbanked individuals. # TrustScore - AI-Powered Alternative Credit Scoring Alternative credit scoring platform for Sub-Saharan Africa. Uses AI to analyze mobile money transactions, telecom data, utility payments, and behavioral patterns to generate credit scores for individuals with no formal banking history. ## Problem Over 350 million adults in Sub-Saharan Africa are unbanked and lack traditional credit histories. Financial institutions cannot assess their creditworthiness, leaving massive revenue potential untapped. ## Solution TrustScore analyzes alternative data sources to generate credit scores (300-850) with full explainability and fairness monitoring. ## Architecture ``` ┌─────────────────────────────────────────────────────┐ │ FastAPI Application │ ├──────────┬──────────┬──────────┬────────────────────┤ │ Auth & │ Data │ ML │ Compliance & │ │ RBAC │ Ingest │ Scoring │ Privacy │ ├──────────┴──────────┴──────────┴────────────────────┤ │ Feature Engineering Pipeline │ ├──────────────────────────────────────────────────────┤ │ PostgreSQL │ Redis │ MLflow │ Celery Worker │ └──────────────┴─────────┴──────────┴──────────────────┘ ``` ## Tech Stack - **API**: Python 3.11, FastAPI, SQLAlchemy 2.0 - **ML**: XGBoost, scikit-learn, SHAP, MLflow - **Database**: PostgreSQL 16 (pgcrypto) - **Cache/Queue**: Redis, Celery - **Infrastructure**: Docker Compose ## Quick Start ### Prerequisites - Docker and Docker Compose ### Setup ```bash # Clone the repository git clone github.com cd trustscore # Copy environment file cp .env.example .env # Build and start services docker-compose up --build -d # Wait for services to be healthy, then seed the database docker-compose exec api python -m scripts.seed_data # Train the initial ML model docker-compose exec api python -m scripts.train_initial_model ``` The API will be available at `localhost: