top of page

The Esther Compiler and Solver

At Global Valuation, we believe in making risk and pricing analytics accessible, easy to use, and fast to compute.

 

Our Esther Compiler and Solver software combination is the result of years of research and development, and has been tested and proven in various areas, including large scale, legal entity-level XVA and margining calculations, model risk, reverse stress testing, and wrong-way-risk. With Esther, quant developers can quickly and easily design and deploy risk and pricing analytics for any use case, and easily integrate them into their existing enterprise solutions and LLM analytics. We are excited to offer this innovative package to our customers.

Please see our video on Risk Analytics in the Age of AI

New video:  Esther XVA model run performance metrics on an Apple M2 Max processor.

Why Esther Works

The Esther architecture is underpinned by an original mathematical framework, providing a multitude of advantages.

Its domain-specific modelling language is capable of addressing all problems in the Pricing and Risk domain in full generality, abstracting away mathematics and low-level model execution from the model developer. This is achieved through decomposition of all mathematical solutions into simple, uniform building blocks that can be combined and executed very efficiently, using vectorisation for parallelising data, and an efficient memory management system with re-use of in-memory data without duplication.

The single, universal solver executes all models as matrix algebra, resulting in very high performance on a small hardware footprint, vertical scalability (falling back to horizontal scaling only in extreme-sized cases), and mathematical consistency in models, data and results across all use cases and instances. With a future-proofed architecture, models are segregated from the execution hardware.

How Esther Works

Esther simplifies the model/market data calibration process, making it reusable for future model runs. After this step, the business logic and portfolio data are formulated into a Directed Acyclic Graph (DAG). The Esther Solver then takes over, performing calculations for any model presented to it by the Compiler.

How Global Valuation Esther works: IDE - Compute - Presentation

By preprocessing input data and storing it in memory along with the solution matrices, Esther Solver is able to quickly traverse the nested Monte Carlo probability tree and calculate the valuations for a vast number of paths.

The analysis stage can be an interactive process where results are stored in memory and further inquiries are addressed through a dialog with the Solver.

How Esther Integrates

Esther is a fully programmable platform structured as a compiler + solver combination.  It is based on language extensions and factors out all the low level coding and mathematical complexities.

Esther is easy to program and also ideally suited to be integrated with AI code generators, such Large Language Models (LLMs).  LLMs provide natural language interfaces and can be taught to produce mathematical solutions for Esther to solve.

Esther software can be used from a web browser, including a Cloud IDE with full debugging and reporting capabilities without any local installation on the user's device.

bottom of page