About Me

Hi there! I'm Phillip Stranger, a passionate computer scientist based in Graz, Austria. Whether you're here to check out my projects, learn about my background, or explore potential collaborations, I'm thrilled to have you visit strangeprojects. Let me tell you a bit about myself and what drives me.

I graduated with distinction from the Graz University of Technology in 2024 with a Master's in Computer Science, focusing on Visual Computing and Data Science. My journey started with a Bachelor's in Computer Science from the same university in 2022, and before that, I completed my A-Levels at Handelsakademie St. Johann in Pongau, which gave me a solid foundation in business administration alongside my tech interests. This blend has helped me approach complex problems with both technical rigor and a practical, big-picture mindset.

Professionally, I've worn a few hats that have shaped my expertise. From 2019 to 2021, I worked as a Web Administrator at Marko Elektrohandels GmbH, where I coordinated the launch and maintenance of a B2B web shop, handled server upkeep, and provided IT support for their ERP system. Then, from 2022 to 2024, I was a Research Assistant at Fraunhofer Austria Research GmbH on the Cross Level User Evaluation (CLUE) project. There, I dove deep into data analysis and developed machine learning models to transform non-expert test results into predictions for experts—super useful for safety-critical systems like air traffic management.

My academic work has been a highlight. My Bachelor's thesis explored simulating LiDAR resolutions in synthetic data to improve 3D object detection, and my Master's thesis introduced a novel way to use non-experts for evaluating expert software. That Master's work even led to a published research paper in the journal Aerospace: "A Novel Approach Using Non-Experts and Transformation Models to Predict the Performance of Experts in A/B Tests" (co-authored with my team and published on July 12, 2024). You can check it out here: https://doi.org/10.3390/aerospace11070574. I've also contributed to open-source projects like Apache SystemDS, which has honed my skills in scalable machine learning systems.

Looking ahead, I'm all about tackling complex challenges at the cutting edge of technology—think data analysis, machine learning, AI, and computer vision. My dream is to work on problems that advance humanity, maybe even in areas like space exploration. If it involves pushing boundaries and making a real impact, count me in! I'm driven, self-motivated, and great at communicating ideas, which helps me thrive in team settings or when overseeing intricate projects.

On a personal note, I've recently gotten into golf, and it's sparked some fun side projects. I'm tinkering with gadgets like a home-built golf simulator and swing analyzer—applying my coding skills to improve my game. It's a great way to unwind and blend my tech passion with hobbies. If you're a fellow golfer or tech enthusiast, I'd love to chat about it!

Thanks for stopping by—feel free to explore my projects or reach out via the contact page. Let's connect and see how we can collaborate on something exciting!

Bachelor Thesis

Title: "LiDAR resolution simulation in synthetic training data for 3D object detection"

Abstract: 3D object detectors need large amounts of annotated training samples to reach high accuracies, but generating these annotated training samples is expensive and laborious. This is the incentive to try to generate these datasets synthetically. Synthetic data describes data taken from virtual environments like game engines. In this work we generate point clouds from the depth images of the Apollo Synthetic dataset to build a dataset on which a 3D object detector is trained. We show the impact the correct simulation of the LiDAR has on the accuracy of the 3D object detector. Additionally we simulate real world effects like noise and drop-out to further increase the accuracy of the 3D object detector.

Master Thesis

Title: "Cross Level User Evaluation"

Abstract: The European Union is committed to modernizing and improving air traffic management (ATM) systems to promote environmentally friendly air travel. However, the safety-critical nature of ATM systems requires rigorous user testing, which is hindered by the scarcity and high cost of readily available air traffic controllers (ATCs). In this thesis, a novel approach is proposed to address this limitation by involving non-expert individuals in the evaluation of expert software, with a focus on air traffic control systems. Using a transformation model incorporating auxiliary information from a newly developed psychological questionnaire, the potential of using students to predict ATC performance with an air traffic control prototype is investigated. Two methods, standard multiple linear regression and quadratic multiple linear regression, were tested for building the transformation model, with both methods enabling high prediction accuracy for the majority of defined performance measures for the ATCs using student test results. However, questionnaire-based metrics showed lower predictive accuracy compared to performance metrics. This study highlights the feasibility of using non-experts for testing expert software, overcoming testing challenges and supporting the User-Centered Design principles.