Modern Computational Finance

This volume belongs on the quantitative implementation shelf next to classical derivatives mathematics, but its subject is the engineering layer that makes large-scale risk run: Algorithmic Adjoint Differentiation (AAD), simulation design, and parallelism. Savine and Andersen write from backgrounds in major dealer quant and systems work; the book’s value is in turning adjoint methods from conference slides into a disciplined development programme rather than a single toy example.

Readers work through how sensitivities can be extracted from pricing code with complexity characteristics suited to enterprise portfolios, where bumping every risk factor is economically and computationally prohibitive. The narrative repeatedly ties back to Monte Carlo and simulation-heavy payoffs, where AAD’s strengths (and footguns) are most visible in production.

The mathematical level is postgraduate quant: readers should be fluent in stochastic calculus at a working level, comfortable reading C++-oriented pseudocode, and familiar with basic numerical analysis. The prose is still more “quant desk + quant dev” than pure probability theory, but this is not an entry-level book.

Risk management, model validation, and front-office quant teams treating XVA, CCR, or desk-wide Greeks as batch problems will find the framing directly relevant: the same computational patterns underpin regulatory stress stacks and intraday risk refreshes. The book helps align quant research, library owners, and IT on why adjoints beat naive bumping when architectures are clean.

Honest scope boundary: the title is not a substitute for a full Monte Carlo theory treatise or a GPU hardware manual; it assumes readers can already price vanilla and exotic instruments at a basic implementation level and now need scalable risk. Where firm-specific scripting languages appear in other Savine volumes, this book stays closer to the core AAD/simulation curriculum.