quant-rate-engine 0.1.1

A high-performance, GPU-accelaerated fixed-income simulation framework.

Rate-Engine: High-Performance Fixed Income & Yield Framework

The "So What?"

In multi-asset portfolios, the ability to rapidly stress-test interest rate sensitivity (Greeks) is often limited by the computational cost of path-dependent simulation. Rate-Engine solves this by leveraging a hardware-aware (MPS/CUDA) stochastic framework that handles 1,000,000+ paths in under 500ms. This enables real-time yield curve calibration and risk-neutral pricing for complex fixed-income derivatives that were previously computationally prohibitive.

Core Features

• Vectorized Nucleus: Built on PyTorch for native GPU acceleration on Apple Silicon (MPS).
• Mean-Reversion Solvers: Optimized Vasicek and Cox-Ingersoll-Ross (CIR) implementations for short-rate modeling.
• Numerical Rigor: Integrated Martingale consistency checks and $1/\sqrt{N}$ convergence validation to ensure unbiased pricing.
• Path-Integral Logic: Vectorized Trapezoidal integration for accurate Zero-Coupon Bond (ZCB) and Discount Factor calculation.

Mathematical Foundation

The framework solves the Ornstein-Uhlenbeck process for the short rate $r_t$: $$dr_t = a(b - r_t)dt + \sigma dW_t$$

And computes the bond price $P(0, T)$ via the expectation of the stochastic integral: $$P(0, T) = \mathbb{E}\left[ \exp\left( -\int_0^T r_s ds \right) \right]$$

Quick Start

from rate_engine.nucleus.sde import Vasicek
from rate_engine.pricing.bonds import BondPricer
from rate_engine.utils.device import get_device

# Initialize Model & Accelerator
device = get_device()
model = Vasicek(a=0.5, b=0.05, sigma=0.02)
pricer = BondPricer(model, device)

# Price 1M paths for a 1-year Bond
price, std_err = pricer.price_zcb(r0=0.03, T=1.0, steps=252, n_paths=1_000_000)
print(f"Fair Value: {price:.6f} +/- {std_err:.6f}")