quant-risk-engine

Python library for curve construction, instrument pricing, stochastic simulation, and XVA, built on QuantLib. Not tied to any specific regulatory framework. Usable as a standalone pricing engine. All data sources are free, no vendor subscription required.

A downstream application implementing EU banking regulatory frameworks (IRRBB, FRTB, CSRBB, liquidity risk) and consuming this library lives in banking-risk.

---

Design

Abstract base classes throughout. Dependency injection. QuantLib evaluation date isolated via _with_eval_date context manager. Curve inputs are typed immutable dataclasses (OISCurveInput, NSSParameters) validated on construction.

---

Curves

OISCurve bootstraps from overnight and par rates across three maturity zones, currently sourced from ECB SDW, swappable to any equivalent input. NSSCurve fits the Nelson-Siegel-Svensson model from ECB parameters. ArrayCurve builds from any externally supplied zero rates.

All three expose the same interface: discount factors, forward rates, and zero rate projection at arbitrary maturities. Used internally by the pricing layer, but also useful standalone for sensitivity analysis and regulatory scenario work: pass any set of tenors and get back interpolated zero rates instantly, ready to apply prescribed shocks.

---

Instruments

Instruments are constructed from their contractual characteristics and priced by passing a curve at call time. The same object reprices against any curve, making scenario analysis and rate bumping straightforward.

Standard Instruments

Priced analytically from a curve. No simulation required.

Bond, IRSwap, FXForward, EquityForward, TotalReturnSwap, VanillaOption.

Credit

• CreditDefaultSwap — flat hazard rate or bootstrapped piecewise curve, CS01, par spread.

Closed-Form Exotics

• DigitalOption — Cash-or-Nothing / Asset-or-Nothing
• BarrierOption — DownOut, DownIn, UpOut, UpIn, analytic Haug/Rubinstein
• ChooserOption — simple chooser, choose call or put at choice date
• CompoundOption — option on an option, Geske analytic engine

Path-Dependent (GBM Monte Carlo)

• AsianOption — arithmetic/geometric, fixed/floating strike
• LookbackOption — fixed/floating strike, continuous monitoring
• CliquetOption — ratchet, sum of forward-start returns
• ShoutOption — lock in intrinsic value once
• NapoleonOption — sum of worst periodic returns
• Accumulator — forced buy at discount with knock-out
• Decumulator — forced sell at premium with knock-out

Multi-Asset (Bivariate GBM Monte Carlo)

• WorstOfOption — worst performer, from_rho / from_prices
• BestOfOption — best performer, from_rho / from_prices

---

Simulation Engine

MCSimulator drives path-dependent and multi-asset pricing. Antithetic sampling, exposure profiles, and stochastic discount factor computation. Supports Vasicek, Hull-White, CIR, GBM, and local volatility processes.

XVA is the applied output of the simulation layer: XVAEngine computes CVA, DVA, FVA, and MVA under netting sets from simulated exposure profiles.

---

Data Sources

Source	Data	Access
ECB SDW	ESTR fixing, MMSR OIS rates, NSS parameters	Free, no key
FRED	SOFR, US Treasury CMT, FX spot rates	Free, API key required
yfinance	Equity prices, ETFs	Free, no key

---

Setup

git clone https://github.com/mrspatbile/quant-risk-engine.git
cd quant-risk-engine
python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"
echo "FRED_API_KEY=your_key_here" > .env
pytest tests/ -v

---

Tests

CI runs on every push and pull request to main via GitHub Actions. Current suite: 460 tests, no live API calls, all fixtures synthetic.