Please describe how you first got in touch with ERGO / ERGO Center of Excellence.

After having applied to the Chair of Mathematical Finance for an applied master thesis in the area of machine learning (ML) in finance, Prof. Dr. Zagst proposed to work on an interesting topic related to this area in cooperation with ERGO.

Furthermore, as I am very interested in machine learning - especially in the application context of insurance - I was additionally invited to join the Machine Learning Team at the Chair of Mathematical Finance for a project conducted in cooperation with ERGO Center of Excellence.

What is your academic background?

After having finished my Bachelor of Science in Mathematics with minor in economics at TUM in 2018, I started my Master in Mathematical Finance and Actuarial Science as part of which I have written above mentioned master thesis. Furthermore, due to my strong interest in Data Science, I have decided to additionally study the Master of Mathematics with a major in statistics and a minor in informatics to gain further theoretical as well as practical knowledge in this area. Both master’s degrees are presumably finished in summer 2021.

What was / is your role within the cooperation between ERGO and TUM?

My role within the cooperation between ERGO and TUM is two-fold. On the one hand, I am a student writing a master thesis which is supervised not only by Prof. Dr. Zagst and PD Dr. Min from TUM but also by Dr. Töpfer from ERGO.  Within the scope of this master thesis, we addressed a real-life problem setting concerning the decomposition of credit spreads proposed by ERGO. To tackle this task, a variety of different machine learning methods as well as a selection of model interpretability techniques were considered and evolved.

On the other hand, I am part of the Machine Learning Team at the Chair of Mathematical Finance in cooperation with ERGO Center of Excellence to tackle a task concerning the improvement of a Generalized Linear Model that is currently used for claim frequency modelling at ERGO. Similar to my master thesis, again a variety of different machine learning methods as well as a selection of explainable AI techniques were considered.

Which experience with ERGO was most remarkable to you?

Particularly during my work in the ML team, I realized that in practice the available data sets of insurance companies are extremely large and not - as it is often the case in lectures - perfectly structured and optimally suited for a certain model. Consequently, preprocessing and modelling is much more complex and time-consuming than expected and might require further resources and techniques to circumvent real-world application problems such as the above-mentioned enormous amount of data, certain data properties like imbalanced data or the requirement of model interpretability.

Furthermore, during the course of my master thesis, I have experienced that interdisciplinary cooperation is often helpful or even necessary, for example, in order to frame, justify and interpret the results obtained when applying the selected methods and techniques.

What do you like best about ERGO and why would you recommend ERGO as an employer?

A very positive aspect that I noticed both within the course of my master thesis as well as during ERGO employees who aim at supporting you.

What is the most interesting thing you have learned from your ERGO supervisor?

Through extensive discussions with Dr. Töpfer, he gave me a feeling of where, how and to what extent Machine Learning might be and already is used in daily business at ERGO.

How will your time working with ERGO influence your future plans?

My collaboration with ERGO has confirmed my ambition to work in the field of data science. In addition, I was able to gain practical experience and knowledge that will certainly benefit me in my future.