title	QuantEngine
emoji	⚡
colorFrom	blue
colorTo	green
sdk	gradio
sdk_version	5.16.0
app_file	app.py
pinned	false
license	mit

⚡ QuantEngine

High-Performance Quantitative Research & Statistical Arbitrage Engine
Accelerated with Apple Silicon Metal Performance Shaders (MPS / NPU), Kalman Filters, Hierarchical Risk Parity (HRP), and GARCH Volatility Modeling.

---

🏛️ Architecture Overview

QuantEngine is a modular algorithmic trading and quantitative analytics framework designed for statistical arbitrage, portfolio optimization, dynamic risk management, and live broker execution.

                     ┌──────────────────────────────────────┐
                     │            Live Market Data          │
                     └──────────────────┬───────────────────┘
                                        │
             ┌──────────────────────────┴──────────────────────────┐
             ▼                                                     ▼
┌──────────────────────────┐                             ┌──────────────────────────┐
│   Statistical Arbitrage  │                             │   Portfolio Allocation   │
├──────────────────────────┤                             ├──────────────────────────┤
│ • Engle-Granger Cointeg  │                             │ • Hierarchical Risk      │
│ • Kalman Filter State    │                             │   Parity (HRP)           │
│ • Ornstein-Uhlenbeck SDE │                             │ • GARCH(1,1) Volatility  │
│ • Metal GPU / NPU ResNet │                             │ • FX Normalized Risk     │
└────────────┬─────────────┘                             └────────────┬─────────────┘
             │                                                        │
             └──────────────────────────┬─────────────────────────────┘
                                        │
                                        ▼
                     ┌──────────────────────────────────────┐
                     │       Broker Bridge & Risk Router    │
                     ├──────────────────────────────────────┤
                     │ • Position Sizing & Slippage Control │
                     │ • Alpaca Markets Execution Bridge    │
                     │ • Live WebSocket Web Dashboard       │
                     └──────────────────────────────────────┘

---

🔬 Core Mathematical & Neural Modules

Module	Core Concept	Description
npu_neural.py	Deep Residual Alpha Network	PyTorch ResNet architecture with LayerNorm, GELU activations, and Bayesian Weight Regularization optimized for Apple Silicon MPS (Metal Performance Shaders) to estimate statistical arbitrage alpha confidence scores.
hrp.py	Hierarchical Risk Parity (HRP)	Marcos López de Prado's tree-clustering portfolio allocation. Uses hierarchical single-linkage clustering, matrix quasi-diagonalization, and recursive bisection to construct robust risk-parity weights without matrix inversion instability.
kalman.py	State-Space Filtering	Adaptive dynamic hedge ratio ($\beta$) tracking. Recursively updates the state transition matrix and measurement covariance to adapt to regime shifts and mean-reversion drift in real time.
cointegration.py	Cointegration & Stationarity	Automated pair selection engine utilizing Augmented Dickey-Fuller (ADF), Engle-Granger two-step cointegration, and Johansen test vector error correction.
garch.py	Conditional Heteroskedasticity	GARCH(1,1) volatility engine for time-varying variance estimation, volatility clustering identification, and dynamic Value-at-Risk (VaR) calculation.
stochastic.py	Stochastic Calculus Models	Ornstein-Uhlenbeck mean-reversion parameter estimation (speed of mean reversion $\theta$, long-term mean $\mu$, volatility $\sigma$), optimal entry/exit threshold solving, and jump-diffusion simulation.
broker_bridge.py	Execution & Risk Management	Production-grade broker integration with Alpaca Markets REST/WebSocket APIs, automated safety caps, and circuit breakers.
web_server.py	Interactive Web Dashboard	FastAPI & WebSocket dashboard providing real-time PnL monitoring, live spread charting, and tear sheet visualizations.

---

🚀 Quickstart

1. Installation

Clone the repository and install the dependencies:

git clone https://github.com/ameerhmz/quant_engine.git
cd quant_engine
pip install -r requirements.txt

2. Running the Engine

Start the quantitative web server and dashboard:

python web_server.py

Open your browser at http://localhost:8000 to monitor live signals, backtests, and portfolio allocations.

---

🛡️ License

MIT License. Developed for research and algorithmic trading experiments.