Masterthesis - AI in goal-based planning

Ortec Finance has developed an advanced goal-based planning methodology, integrated in both a calculation engine available as an API solution, and an interface for wealth & financial planning. This goal-based methodology projects a client’s current aggregated financial situation (including investment portfolios) towards the future and links assets to financial goals. This projection is based on our proprietary economic scenarios. This projection is used to calculate the feasibility of financial goals and typically to reach an ‘optimum’ financial advice through what-if adjustments to for example a client’s investment strategy or periodic deposit.

A common problem is the optimization of steering variables (e.g. periodic deposit, investment profile) to attain a target feasibility of financial goals, taking into account the fact that multiple, conflicting goals with differing priorities might exist and that there are numerous steering variables available. This multi-objective, multi-variable  optimization problem is furthermore constrained by performance (calculation times need to be fast) and advisor demands (needs to be explainable & take into account client preferences regarding steering variables).

With the recent rise of AI & machine learning techniques, we are currently investigating approximating the calculation engine, i.e. creating a meta-model by parametrizing input and output parameters and approximating the relation between these. This serves to the end of reducing the computational time for calculations, in order to speed up optimization problems.

One possible approach in scope is the use of a neural network to model the calculation engine, bootstrapping the data required to feed into training. Subsequently, this model can be used for rapid optimization across multiple steering variables.

The aim of this project is to:

  • Assessment of appropriate techniques for meta-modelling and optimization based on the calculation engine;
  • Develop a meta-model of the simulation engine, parametrizing input and output parameters, to approximate it to a high degree of accuracy, in predicting e.g. goal feasibilities;
  • Bootstrapping simulated data to feed into the model in order to train & test it;
  • Using the neural network to tackle multi-goal, multi-variable optimization problems, incorporating imposed restrictions on the optimization domain;
  • Assessment & validation of the feasibility, accuracy and performance gains of such a model, compared to the calculation engine;
  • Comparison of the model to benchmarks in optimization problems.

Requirements for this project are:

  • Interest in the application of quantitative methods in finance
  • Solid background in mathematics
  • Knowledge of machine learning methodologies (e.g. deep learning) and tools (e.g. Keras, Tensorflow)
  • Experience with scientific programming languages such as Python

References

Janssen, Kramer, Boender (2013) - Life Cycle Investing: From Target-Date to Goal-Based Investing
Zhao (2018) - Neural network copula portfolio optimization for ETFs

What we need from you

To take your application in to account, we would like to receive several documents from you:

  • An up to date CV
  • Your motivation for Ortec Finance and this assignment
  • Information about the requirements in terms of start date, period etc.
  • List of marks

More information or to apply?

For more information regarding this position please contact one of our HR colleagues at +31 10 700 50 00. To apply, please use the application button below to send us your cover letter, CV and list of marks.

We hope to meet you soon!

Apply for this vacancy