Magnus-Archon/Investor-Founder-Matching-System

Multi-model hybrid ML system for Investor-Founder Matching for Nigerian/African startup ecosystem. # 🇳🇬 Investor-Founder Matching System Multi-model hybrid ML system for Nigerian/African startup ecosystem. ## Quick Start ```bash pip install -r requirements.txt python3 scripts/train.py # retrain all models uvicorn api.main:app --port 8002 # start API ``` API docs: localhost ## Models - GradientBoosting (Optuna 15 trials) — val AUC ~0.905 - RandomForest (Optuna 10 trials) - Logistic Regression (L2) - Neural Bilinear Two-Tower (PyTorch, 150 epochs) - Startup + Investor Autoencoders (16-dim bottleneck) - TF-IDF + LSA text embeddings (64-dim joint vocab) - Optuna ensemble blend (30 trials) - MLflow experiment tracking (models/mlruns/) ## Data 60 real African startups × 104 real investors Sources: Partech Africa 2024, TechCabal, Disrupt Africa, Techcrunch, press releases ## Key Endpoints POST /match/new-startup — match any startup against all investors GET /match/startup/{id} — top investors for a listed startup GET /match/investor/{id} — top startups for an investor POST /match/score — score a specific pair GET /analytics/summary — ecosystem overview GET /analytics/top-matches — highest scoring pairs globally GET /model/feature-importance — what the model learned