native speaker.
C2 proficiency.
my main language; used across all my AI/CV projects and my thesis.
my go-to framework for deep-learning research.
used heavily across 5 years at ATX Hardware and in university projects.
rendering work, my bachelor thesis, and private projects.
custom shaders in private and university projects, and at work.
used frequently in private and university projects.
working knowledge from coursework and projects.
used in private projects (including this site).
everything I do is, and always has been, very organized.
practice makes perfect.
acquired by leading many university projects.
🌾+🌳=🍻
🥖🥐🍕
🥩+💨=🍖
🎸+🙋♂️=🤘🏻
🏰+🐉=💰
I'm passionate about building robust, readable, and well-structured software, and I love turning hard technical problems into clean solutions. My focus is artificial intelligence and computer vision, currently as a master's student at Ulm University, where I work on self-supervised learning, vision-language-action models, and generative models.
A few recent highlights: my paper Bar-JEPA was accepted at ICDAR 2026 (first author), I'm maintaining a strong grade average (≈ 1.2), and I stay close to current research through a weekly paper-reading group with PhD students. I learn fast, care about doing things properly, and am always chasing the next interesting problem.
Focus on computer vision, deep learning, and generative models. Current grade average ≈ 1.2. Project work in self-supervised learning, vision-language-action models, and model-based reinforcement learning.
Transferred to the Artificial Intelligence M.Sc. at Ulm.
Thesis (graded 1.0): Photorealistic Rendering of Training Data for Object Detection and Pose Estimation. Studies paused 2015–2017 for a full-time engineering role at ATX Hardware.
EQF Niveau 4
Designed, built, and programmed ESP8266-based IoT devices, including a beverage-fermentation/keg monitor and a wallbox power-automation controller. Self-taught embedded electronics and 3D-printed hardware design; kept this portfolio site running.
Replaced the NH-Software legacy GDI+ software-rendering path with a GPU-accelerated 2D engine (heavily modified SFML, C#/C++, custom shaders), eliminating multi-second UI freezes and keeping large circuit-board projects responsive in daily use. Added Line-Arc DXF geometry support and a marching-squares outline-extraction algorithm that converts safety-overlay regions into precise line/arc vector geometry. Also built control software for the "Lomecs" laser-guided optomechanical imaging machine (VB.NET, serial motion control) and a Raspberry-Pi camera system.
Developed and maintained in-house .NET software for circuit-board test-adapter design, used daily by the sales and engineering teams.
Weekly ML paper discussions with PhD students.
MIT 6.S184: Self-study labs.
