FinRisk.io is a next-generation financial simulation and monitoring platform designed to quantify, visualize, and predict systemic contagion in interbank networks. By blending advanced network topology, predictive machine learning, and behavioral game theory, FinRisk.io offers a comprehensive "Stress Test as a Service" for modern financial ecosystems.
FinRisk.io is built with a "Cyber-Industrial" aesthetic—modern, clean, and highly interactive.
- Dynamic Glassmorphism: Translucent panels and blur effects for a premium dashboard feel.
- Vibrant Visuals: Emerald-to-Rose gradients indicating risk levels from healthy to hazardous.
- Micro-animations: Smooth counting animations, pulsed risk alerts, and interactive graph flows.
- User-Centric Navigation: A persistent sidebar providing instant access to deep analytics modules.
- Interactive Graph View: Visualize millions in interbank exposures using
ReactFlow. - Local & Global Views: Toggle between high-level systemic overviews and node-specific "Local Neighborhood" views.
- Real-time Modifiers: Inject liquidity or simulate partial defaults with instant visual feedback.
- Cascade Visualization: Watch defaulted debt propagate through the system in real-time.
- Spectral Stability: Monitor the Systemic Risk Score (Spectral Radius
λ_max). - Timeline Replay: "Time Travel" through simulation steps to identify the precise moment of systemic failure.
- Inference Intelligence: A Random Forest regressor analyzes 10+ topological features to predict collapse probability.
- Feature Importance: Understand why the system is at risk (e.g., "Node Degree Variance" vs "Density").
- Forecast Horizon: Predictive analytics for T+1 to T+3 simulation steps.
- Realistic Market Opacity: Simulate environments where only direct counterparties are visible.
- Shadow Node Logic: Backend-driven data masking ensures true information asymmetry for stress-testing transparency.
- Strategic Analysis: Solve for Nash Equilibrium in the interbank lending game.
- Payoff Optimization: Analyze how individual bank behaviors (e.g., hoarding liquidity) impact total system welfare.
- Strategic Explanations: Gemini AI explains complex graph metrics in human-readable terms.
- Policy Advisor: LLM-driven regulatory advice based on current systemic stress levels.
- React 19 & TypeScript: Typed, high-performance UI foundation.
- Vite: Lightning-fast build and development environment.
- Zustand: Lightweight, high-performance state management for simulation data.
- Tailwind CSS: Utility-first styling with custom themes.
- ReactFlow: Powerful node-based graph editor for network visualization.
- Framer Motion: Fluid UI transitions and micro-animations.
- FastAPI & Pydantic: Modern, high-performance Python web framework.
- NetworkX: Industrial-grade graph algorithms and topology analysis.
- NumPy & Scikit-learn: Matrix math foundation and ML forecasting models.
- Google Generative AI: Powering the AI insights and policy guidance.
- Python 3.10+ (Conda or Venv recommended)
- Node.js 18+
- npm (or yarn/pnpm)
cd backend
python -m venv venv
# Activation
# Windows:
.\venv\Scripts\activate
# Unix/MacOS:
# source venv/bin/activate
pip install -r requirements.txt
uvicorn app.main:app --reloadNote: Ensure you have configured your
.envwith aGOOGLE_API_KEYfor AI features.
cd frontend
npm install
npm run devThe application will be live at: http://localhost:5173
├── backend/app/
│ ├── features/
│ │ ├── network/ # Graph generation & logic
│ │ ├── propagation/ # Shock contagious logic
│ │ ├── risk_matrix/ # Heatmap & Var calculations
│ │ ├── ml_engine/ # Scikit-learn models
│ │ └── ai/ # Gemini LLM integration
│ └── core/ # Global configs & math foundations
│
├── frontend/src/
│ ├── pages/ # Layout components (Dashboard, ML, etc.)
│ ├── components/ # Specialized UI modules (Cards, Graphs)
│ ├── store/ # Zustand simulation state
│ └── services/ # Fetch-based API client
Developed for the CodeSib DataThon.
"Quantifying fragility, predicting stability."