64 job analyses later: why I would rather build than talk

I currently have 207 active jobs in my pipeline. I have looked at 64 of them in detail. Only now does the actual applying start.

That does not sound like progress at first. For me it was exactly that.

Some context: I take an interest in far too many topics. Given the choice I would try my way through every tech and AI related job there is. The only problem: my degree (Management, Communications, IT) gave us the basics of programming, but not as deep as I would sometimes like.

It feels like every second posting asks for Python, know-how about vector databases, chunking and the rest of it.

This text sits in chunks

So here we are. You can talk to my chatbot on this site now. While building it I simply had the topics explained to me as I went: why do I split my texts into chunks? How do I keep the system safe so a single user cannot burn all my tokens in one session? In the end that turns the theoretical requirements into something I can actually apply.

The text you are reading right now lives in Supabase, by the way. You should be somewhere between the first and the second chunk. When you write to the chatbot, the model compares the patterns of this chunk against what your prompt asks for. On a match it uses this passage as context instead of hunting for specific words with plain SQL queries. And in case you are trying to drain all my tokens: good luck, I use Upstash for rate limiting. ;)

Business meets code

Now that building apps and software has become noticeably easier, I can make up for my gaps in pure programming rather well. And suddenly the rest of the degree pays off too: the basics of business administration, law, process and project management, marketing, and above all the practical projects run directly with companies.

I can spot where a system actually moves us forward and supports my own processes. And I understand how much the technical and legal safeguards matter. For that business knowledge I now only have to find the right tool or the fitting technical solution, and that only is doing a lot of work in this sentence.

Those learnings stick. Within a few hours, right on the project. My main focus at the moment is my bachelor thesis after all. The rest are late night projects.

Projects instead of applications

That is why I bet on tangible projects instead of mass applications.

I do not want to walk into an interview and say: yes, I heard something about RAG architectures and APIs at university once. I want to say: have a look at kilianwarmdt.de. See if you can trick the chatbot.

Another example is my career pipeline web app. 207 active jobs, 64 roles evaluated in detail. I would never have kept that up in Notion. So I built my own system: Claude compares my profile directly against the postings and files them in my app. From there I move them into the next stages like in a CRM. My Obsidian vault updates at the same time. That way my AI always has the current state and I have all the data ready for a deep analysis.

That is what I value about building things: I learn a lot in a short time and fix my own everyday problems while doing it.

And maybe it is a real advantage in a job market that is not kind to young people right now. Everyone claims they can work with AI. I would rather show it with my own examples.