We Did it Right by Doing it Wrong?

Technical and economic feasibility do not always go as a couple. Since we launched UnRisk, we made decisions that might have seemed questionable at the time they were made. Design principles, algorithmic innovation, transparency and openness  ... do not always pay back immediately.

REINVENT THE FOUNDATION - when we entered into the quantitative finance market, tree-based and Montecarlo methods were widely propagated. 
We transferred numerical schemes from complex technical system solving to finance:
  • Methods for the stable solution of partial differential equations. Keywords here are finite elements, asymptotic math approaches, and techniques for handling convection-dominated equations (upwinding, streamline diffusion).
  • Methods for robust calibration. As the parameter identification problems in finance are typically ill posed, one should know about the regularization techniques which make ill-conditioned problems tractable. If the forward problem has no closed-form solution, adjoint method techniques are used.
  • FFT and wavelet techniques.
  • and clearly (Quasi)Montecarlo with Longstaff Schwartz ...
The choice of the right solvers is the very kernel.  They enabled us to extend the coverage of deal types unprecedented quickly. In risk management, where numbers are crunched many millions of time, they must be accurate and robust. 
And even now, having enormous computing muscles, allowing for algorithmic simplification they are indispensable benchmarks in model validation.

USE MIXED LANGUAGE CONCEPTS - when we started, not only solvers were written in C++. The learning and interaction effort for teams is minimal, when having a one-language development environment. We decided to integrate our number crunchers into Mathematica, not common in financial circles at that time. We exploited two major principles of Mathematica:
  • a declarative programming environment
  • link technologies that allow the integration of data bases and other languages, like Java, Python, ...
We created a domain specific language enabling us to develop parts of UnRisk in UnRisk and built our UnRisk FACTORY that combines valuation and data management controlled by Java services and SQL, and webbed-it by web services, Java server pages, ... 
This house stands stable, although no implementation stone is left unturned  (UnRisk Goes Cross Platform) right now. Driving OpenCL over grids.

UNLEASHING THE INNOVATION - making UnRisk-Q, our own development environment, available to quant developers we reinvented our business. It was not an obvious decision. But to maintain it without compromise, we have established the UnRisk Academy completing the tools to know-how packages,

We want to do where we are good at. And serve customers first, whilst driving generic technologies. We rather want to partner and run a company without borders than growing by high pressure strategies ....