Technology Research and Development
We build models with passion
Our aim
Think along with the clients and offer them the convenience they need
Very often it is very expensive for clients to hire consultants on-site to develop a model from scratch, suppressed by internal or external deadlines. We aim to provide a much more economic solution for them, especially when the budget is a problem or the time pressure is high.
Our solution is to build calculation kernels and tools up front, so that, with those readily available and carefully tested, the clients only need to finish the rest of the model development cycle themselves or with our help. As such, a considerable amount of model development time and cost are saved.
In case the clients do need a complete re-implementation in their own systems, our calculation kernels and tools can be used as blueprints or prototypes, again saving considerable amount of development time.
Our readily available calculation kernels and tools can be used for independent validation purposes, too.
Our team has a broad coverage of skill sets necessary for research and tooling business, from model and algorithm development to prototyping and implementation on production level. Having a scientific yet pragmatic attitude is what we are particularly proud of. Further, we constantly engage ourselves in doing research on advanced models/methods and cutting-edge applications of new technics (such as machine learning methods) in financial industry.
For the moment, we have one tool ready for sale and one calculation kernel ready for demo. A few other tools are in the development pipeline. Please do feel free to contact us, in case you cannot find what you are looking for from below. In the background, usually in terms of Msc internship projects, we have more models being developed. We can take look at your specific need and let you know what we can do.
Progress
We are confident to do better
Existing Achievement
Derivatives pricing and financial risk calculation based on the Fourier method (that is to continue the COS method originally developed by the founder Dr. Fang Fang; this method has been widely adopted in the industry for more than 10 years) and combined with machine learning technology.
Current Research
High-dimensional numerical methods and financial applications; interpretable deep neural networks and financial applications
Representative Papers and Recent Papers
1. A novel pricing method for European options based on Fourier-cosine series expansions, Fang Fang and Kees. Oosterlee, SIAM J on Scientific Computing, 31(2), 826-848, 2009.
2. A new and efficient Fourier method for risk quantification and allocation of credit portfolio, Xiaoyu Shen and Fang Fang, conference paper presented at Bachelier World Congress 2022 and SIAM conference on Financial Mathematics and Engineering 2023.
3. Fast calculation of Potential Future Exposure and XVA sensitivities using Fourier Series Expansion, Gijs Mast, Xiaoyu Shen and Fang Fang, conference paper presented at SIAM conference on Financial Mathematics and Engineering 2023.
4. A novel solution method for multivariate expectation problems based on dimension-reduced Fourier-cosine series expansion, Fang Fang, Marnix Brands, and Xiaoyu Shen, conference paper presented at SIAM conference on Financial Mathematics and Engineering 2023.
Tools for sale
We guarantee quality and accuracy
FRTB-SA Calculation Tool
This tool contains a complete calculation package of FRTB-SA: SBM (sensitivity based method), DRC (default risk charge) and RRAO (residual risk add-on).
We could either embed the readily-available calculation kernel into a client’s existing system or re-implement the full calculation based on our prototype.
We also provide associated consulting services, such as composing functional specifications for FRTB sensitivities and unit testing the calculation results
For clients who have their own or vendor calculations ready, we could help with a fast validation using our tool as benchmark.
Ready for demo
Tested and readily available; Can be used for multiple applications
Calculation Kernel for Risk Measures and Sensitivities of Portfolio Losses under Copula-Factor Model
Model: Multi-factor Gaussian copula, Multi-factor t copula, Multi-factor Gaussian-t hybrid
Method: Monte Carlo simulation and COS method; The COS method application in this field is noval, from our own research. A paper on this is to be submitted for publication and will soon be available online.
Applications: IFRS9 for CDOs (such as SME senior tranches); Economic capital for banking book loans; FRTB-IMA-DRC; IRC under Basel 2.5.
Tools in the pipeline
A fast calculator of PFE, EPE, ENE, EEPE, etc. on counterparty or netting set level, and XVA sensitivities
The COS method will be used to replace the time-consuming Monte Carlo simulation
This enables practitioners to install a small calculator on their local pc for quick and easy pre-trade assessments, sensitivity or ad-hoc analysis. A paper on this is currently being drafted for publication.
Data Back-filling and prediction based on machine learning methods
Conventional method: Principle component regression
New method: Various machine learning methods
Predictor of the Maximum Drawdown of a Portfolio
Simulation of the maximum drawdown based on multi-factor copula
CDS Pricer & Bootstrapper
Simple model and ISDA model
Other: tailored to your needs
We have more models that are being developed in the background. Feel free to contact us in case you are looking for some particular calculation engine or model. We are always ready to develop a prototype or a tool that fulfills your needs and standards.
Theses Presentation
7. A novel pricing method for European options based on Fourier-cosine series expansions