Sentinel
Financial Intelligence Platform
Real-time data ingestion, anomaly detection, time-series forecasting, and news-sentiment analytics — served through a Bloomberg-style React dashboard with a FastAPI backend.
capabilities
What it does
7 Dashboard Pages
Market Overview, Anomalies, Forecasts, Technical Analysis, Fundamental Analysis, Portfolio, and Sentiment — all in a Bloomberg-style React SPA.
Anomaly Detection
Isolation Forest + LOF ensemble from PyOD, with severity labels and on-demand detection via API.
Forecasting
Prophet + XGBoost blended 50/50 with walk-forward cross-validation at 30/90-day horizons.
Portfolio Optimization
MPT, Black-Litterman, Kelly Criterion, and Blended allocation methods with efficient frontier visualization.
Drift Detection
PSI + KS tests on 11 features against a 180-day baseline, with email and Slack alerts.
CLI
stl ingest, stl ml-train, stl api-start — or stl all to run everything in sequence.
dashboard
7 React pages
Bloomberg-style dark theme with Plotly.js charts, lazy-loaded via React.lazy().
architecture
How it's built
┌──────────────────────────────┐
│ FastAPI :7860 │
│ /api/v1/* → routers │
│ /api/docs → Swagger UI │
└──────────────┬───────────────┘
│
┌──────────────▼───────────────┐
│ PostgreSQL (SQLAlchemy) │
│ Supabase / local / managed │
└──────────────────────────────┘
Frontend: React 19 + Vite + Plotly.js (Render)
ML: PyOD · Prophet · XGBoost · FinBERT · MLflowstack