(2023-2024) Modernization and further development of a web app for the PropTech sector
The Brick Cloud enables B2B customers of Brick4u GmbH to visualize and export measurement data from electricity, water, gas meters, etc.. There is also an interface to the so-called "brick" devices that collect the data from the respective buildings in the first place. Customers can conveniently configure the measurement intervals and additional parameters from the web app.
(2022-2023) Development of a web tool for test automation in the banking industry
In this project, I worked in an international team to realize a full-fledged test automation software that offers all the features that the customer, a large European payment service provider, wants. Since there are many different data formats in the banking industry and an affinity for detail, it was important that the software takes all these circumstances into account and offers system testers the possibility to create and run test cases specifically tailored to banks. The software was implemented on the web with Angular in the forntend and Python in the backend.
(2021-2022) Pilot Process Mining Case Study
The regional social welfare oranizations "Landschaftsverband Westfalen-Lippe (LWL)" and "Landschaftsverband Rheinland (LVR)" are responsible for inclusion of disabled people and the distribution of public funds. In a cooperation project with the goal to invesitage the impact of Process Mining, over 50,000 business cases were provided. Next to problems regarding the data quality, also the need for specific domain knowledge was a challenge to deal with. Nevertheless, the process data was successfully analyzed, especially the conformance between the should-be-process and the as-is-process was determined. Along the way, a new scientific method was proposed and developed within a prototype which is able to detect and visualize deviations from the desired process directly in BPMN models. The cooperation partner was very impressed and will make use of Process Mining in further projects.
(2019) Vehicle parameter based prediction of lap times
This project had the clear goal to develop an algorithm which can predict the ideal lap time of a car given its parameters. There were plenty of difficulties to deal with like the danger that the model would rely too much on the training data and could not perform well on unseen data. In addition to that, I had to find a way to make the prediction reliable for new race tracks because the data I was given was based on one single race track. I developed several models and compared them with each other. The result was that an Adaboost regressor worked way more better than neuronal network based approaches when dealing with new data. At the end, my model performed well in the tests and the customer was satisfied.
(2018) Interactive sports analytics tool for comparing football sequences based on running trajectories
In this interdisciplinary research project, the aim was the development of an interactive tool which helps sports analysts to detect interesting scenes and tactical patterns faster. I designed a desktop application where an analyst can load raw position data from football games. These data are generated from sophisticated camera systems in football arenas. When the data enters the program, they become preprocessed. Afterwards, it is possible to select scenes that are separated from each other by ball possession. Now, the analyst can choose several player trajectories of the current scene and start a retrieval procedure which returns the most similar other scenes to the current scene. The similarity is based on trajectory alignments. A coordinated evaluation in cooperation with sports scientists showed that the tool is indeed useful for the investigation of tactical patterns.