Between classical engineering and new tools.
What interests me about AI is not only how much faster some things become, but what becomes possible when you bring a solid technical foundation and use these tools deliberately.
As a millennial, I still grew up with sandboxes and corded phones, then went through the entire digital shift from computers and smartphones to today's language models. I am genuinely glad about that mix: I learned to program the classical way, built a solid technical foundation, and still stayed open to the next big leap. As a tech nerd, the AI wave caught me early for exactly that reason. Not just as a spectator, but with real excitement about what suddenly became possible.
For me, the value of AI is not about giving up the thinking. It takes on a lot of the heavy lifting so I can focus more on the part I actually enjoy: understanding problems in detail and developing elegant solutions from there. You can see that on this website as well. Two languages, different modes, and a lot of structural detail would have taken much more time before. Today I can move much faster from an idea to a working solution without losing my overview of the system.
ChatGPT appeared while I was doing my master's degree. I still remember those first prompts and how immediately compelling that felt. The topic then became part of my master's thesis as well, but not in the sense of having something quickly hallucinated together. What interested me was how large language models such as GPT-4 could be used to generate synthetic training data for machine-learning applications in healthcare.
That question is especially interesting there because good models need good data, while privacy, sensitive information, and small datasets are often a real obstacle. My work showed that synthetic data can improve model performance when used as an addition, especially in areas where real data is scarce. At the same time, it also became clear that synthetic data cannot simply replace real data. That mix of potential and limitation is exactly what still makes the topic interesting to me today.
Today I mainly use AI as a tool for thinking, structuring, programming, and writing. Especially for text, that is a real advantage for me: I can first capture or dictate ideas, knowledge, and thoughts directly and then turn them into a useful contribution step by step.
This is where the projects and articles sit in which AI is not just a quick gimmick for me, but a tool with actual value, clear limits, and a concrete use case.
An overview of my master's thesis on how large language models such as GPT-4 can augment synthetic training data for machine-learning applications in healthcare.
A project at the intersection of AI, small tools, and 3D printing, where abstracted aircraft CAD data eventually turns into physical cookie cutters.
A look at how I used AI while building this website, where it was extremely strong, and where you still have to think cleanly yourself.