Rust · Python · CUDA · Metal
stochastic-rs
High-performance Rust library for stochastic process simulation, option pricing, calibration, statistical estimators, copulas, and neural-network volatility surrogates. 120+ processes, 19 SIMD distributions, and a typed pricing engine — with first-class Python bindings.
120+ processes
Diffusion, jump, fractional, rough, short-rate, HJM, LMM. Each with a generic-precision ProcessExt<T> impl and CUDA / SIMD acceleration where it makes a difference.
Pricing & calibration
BSM, Bachelier, Heston, SABR, Bergomi, rough Bergomi, double-Heston, CGMYSV, Hull-White swaption. Closed-form, Fourier, Monte Carlo, finite difference, lattice, Bermudan LSM.
Statistics & estimators
Hurst (Fukasawa, R/S, …), MLE for 1-D diffusions with 6 transition-density approximations, ADF / KPSS / Phillips-Perron, realised variance with BNHLS bandwidth.
Risk & credit
VaR / CVaR / expected shortfall, Sharpe / Sortino / Calmar with no hard-coded annualisation, Merton structural model, hazard bootstrap, JLT migration with Padé-13 matrix exp.
Vol surface & SVI
Implied-vol surfaces from market quotes, SVI / SSVI / SABR-smile fits with arbitrage-free interpolation. Plug a model into ImpliedVolSurface::from_flat_iv_grid for fast inversion.
Python bindings
Full coverage via PyO3 — 198 classes + 12 functions across distributions, processes, pricers, calibrators, copulas, and stats. NumPy in / NumPy out.