• 💻 DataScientist
• 🚴 Cyclist,👨‍🚒 Firefighter,⛵ Sailer

• Successful initiative for digital forest fire prevention in Germany
• A free and public API to inform the German citizens about a high risk of forest fires and preventing forest fires caused by negligence
• It uses official data and provides a nice and easy way to access the data anywhere
• 10k+ daily unique API calls
• WBS API

• Using Attention Techniques for Explainability of Deep Learning Models in Computer Vision
• My bachelor thesis presents a novel explainer for attention-based vision transformer, which visualizes and explains the results of image classification tasks. The results are measured by quantitative metrics and compared against the established LIME explainer. Although the attention-based explainer performed slightly worse, further fine-tuning can address the identified issues. Overall, this work contributes to enhancing comprehensibility and trust in AI models by combining attention mechanisms with model explanations.
• Link to Thesis, Code and Presentation

• Synthetic data crafted for educational and scientific purposes — a collaborative effort by five students. Over the course of our collaboration, we tackled the prevalent challenge faced by educational institutions in securing reliable datasets suitable for academic lectures and research initiatives. Our focus has been on developing algorithms capable of generating such datasets. Importantly, these datasets exhibit discernible patterns, providing ample opportunities for diverse and sophisticated research inquiries across various subject areas, while ensuring contemporary relevance.
• Our project centers around the creation of master and transaction data for an online shop affiliated with a fictional candy manufacturer. Leveraging authentic statistical distributions and an Agent-based Modelling simulation, our algorithms simulate the impact of products and advertisements on customer orders. This ensures the generation of datasets with over 45 distinct patterns, enabling students and experts to conduct analyses through visualizations and contemporary statistical methods from machine learning. As an illustration, our data indicates a higher demand for cough drops during winter compared to summer.
• In summary, our project stands as a notable success in terms of planning, execution, and outcomes. We effectively addressed practical challenges using modern technologies and are delighted to present our project to the Nordakademie for integration into lectures and research endeavors.

• Identifying Ultimate Beneficial Ownership from Complex Corporate Structures using Graph Machine Learning
• My master's thesis introduces an innovative graph machine learning methodology designed to identify ultimate beneficial owners within complex corporate structures. This research holds profound significance for regulatory authorities relying on such analytics to combat illegal activities like money laundering and tax evasion. The model is quantitatively evaluated and compared to the established algorithm across various general and problem-specific metrics. While the trained model demonstrated marginally inferior performance in identifying ultimate beneficial owners, its true strength emerged in addressing key challenges inherent to the algorithmic approach. It exhibited significantly improved runtime efficiency when dealing with deep and complex corporate structures. Moreover, it excelled in scenarios where the algorithm failed to yield any solution. In summary, the findings of my thesis represent a substantial step forward in the analysis of complex corporate ownership structures, offering a more efficient approach. This research enhances the effectiveness of regulatory efforts and promotes transparency within the corporate landscape.
• Link to Thesis and Code

From the very beginning, my education was characterized by a great interest in mathematics, algorithms and understanding how the world works. With the excellent combination of theory and hands-on experience, I have always been able to gain important experience and challenge and improve myself every time.

• Courses in machine learning, text analytics, process mining, Mining, Statistics, Data Mining and much more
• Thesis: Identifying Ultimate Beneficial Ownership from Complex Corporate Structures using Graph Machine Learning
• Final grade: 1.3

• Courses in software engineering, databases, basics of AI, mathematics and much more.
• Thesis: Using Attention Techniques for Explainability of Deep Learning Models in Computer Vision
• Final grade: 1.7

• Final grade: 1.3