A train out of Hamburg
Some mornings the train from Hamburg carries you further than you booked for.
One summer, I got on one with two friends, a backpack, and a book someone had handed me the week before. I hadn't opened it yet. We were going to the Vysoké Tatry — the High Tatras, the line of mountains where Slovakia presses up against Poland — for ten days of hiking. I had no plan more elaborate than that. I was a Financial Accountant ("Bilanzbuchhalter") at Otto Group and had been for almost five years. Books for several subsidiaries, monthly and quarterly closes under HGB and IFRS, UiPath bots I'd helped build to automate vacation accruals. A career quietly on the rails.
Ten days of walking
The rhythm of those days was very simple. Train from Hamburg to Poprad. A small village at the foot of a valley. Walk all day from one place to the next — uphill until your legs complained, downhill into another village, another small guesthouse, another shower that felt like a small miracle. Dinner. A beer. Cards. Laughing about nothing in particular. A walk through whatever cobbled square the village had. And before sleep, a few pages of the book.
The book was Café am Rand der Welt by John Strelecky. Three questions sit at the middle of it like three small stones in a pocket: Why am I here? What do I want? Am I living it? Nothing original about the questions. But on a long mountain walk, with nothing to distract you but the weight of your own boots and a 45-liter backpack, they stop being abstract and start becoming a little dangerous.
The long hikes gave us two kinds of time. Long stretches of talking — the kind of conversation that only happens when people are looking at the same trail and not at each other's faces. And longer stretches of quiet, especially on the steep parts, where nobody has breath to spare for words. The quiet is where the questions got loud.
The trail between Starý Smokovec and Štrbské Pleso
I still remember the long route between Starý Smokovec and Štrbské Pleso — such a magical place, and an amazing route to reach. It's one of those walks where the trail keeps opening onto a new valley, and you keep thinking you've arrived, and you haven't. You just keep walking.
Somewhere on that stretch, I wasn't thinking about work. I was thinking about the last time I had built something for the joy of it. Playing with tech, tinkering with small things, bringing an idea on a napkin to something that comes alive — a product, a small project, a toy that finally clicks together. I hadn't done that in years. Life had drifted, quietly and without asking, away from the thing I'd always loved most. Bookkeeping is a fine craft. It's just not the one I came for.
On that trail, between one mountain lake and the next, I made a decision. I didn't announce it. I didn't tell my friends at dinner. I just knew. The job was done. Whatever was next, it wasn't this.
Back in Hamburg
When I got home, I quit. The plan at the time was softer than it sounds: continue my studies, do a Master's, find my way back to something closer to my old passion. I thought about Wirtschaftsinformatik — business informatics. A quiet path back to tech, through a door I already half-knew.
Then one evening, while researching programs, I stumbled into the story of AlphaGo.
DeepMind's machine playing the ancient game of Go against Lee Sedol, nine-dan, one of the strongest Go players in history. I read about the architecture, the Monte Carlo tree search blended with deep neural networks, the way a few hand-coded heuristics had given way to something that could teach itself. I read about move 37 in game two — the move no human would have made, the move that made experts sit forward in their chairs. It was the first time I understood that intelligence itself was being engineered.
I fell in love with AI the way people fall for a piece of music. Not as a career choice. As a direction.
The unglamorous route
From there the road was longer and less cinematic than I'd like to pretend.
In November 2022 I started a 540-hour Data Science bootcamp at neuefische in Hamburg. I should be honest — I didn't already know Python. The first two weeks humbled me completely. Statistics, machine learning, deep learning, SQL, data engineering, visualization: they came at us in waves, and I spent evenings catching up on the mornings. The capstone was a multi-objective recommender system for the OTTO Kaggle challenge — Word2Vec embeddings, co-visitation matrices, the whole thing running on GCP.
The following year was quieter and harder. Between bootcamps I built small projects and passed the Telc C1+ German exam to apply for the master's. I applied for a lot of jobs and heard back from almost none.
Then in early 2024 I enrolled in a second neuefische bootcamp — 540 hours of full-stack Java development. Spring Boot, OOP, MongoDB, OAuth, Docker, TypeScript, advanced React. I wanted to be able to ship end-to-end, not just deliver a notebook. The capstone was MyRecipes — an AI-powered recipe app that's still quietly running on Render today.
In April 2024 I started my M.Sc. in Data Science & AI at FH Wedel. Deep Learning. Computer Vision. Econometrics. Symbolic AI. Category Management, E-Commerce, Digital Transformation — the business track, because I wanted the two halves of my life to talk to each other properly.
Then the Future Founder program — workshops, coaching sessions, and pitch events to develop a startup mindset. In a team of four, we built an AI-powered regulatory compliance tool for biotech — streamlining processes to accelerate life-saving products to market.
Then Datalogue as an AI & Data Science working student, where I shipped the B2B sales lead pipeline (a story of its own). Now the Master's thesis, on Document Intelligence and Knowledge Graph Construction — OCR-free extraction, knowledge graphs over document corpora, GraphRAG reasoning with local LLMs.
What finance gave me
I was worried, in the Tatra mountains, that five years at an accounting desk would be a weight around my ankles. They aren't. They are the most useful thing on my CV.
You learn what stakeholders actually need, not what they ask for. You learn to explain a difficult number to someone who has to sign it off. You learn how enterprise systems actually breathe — SAP, spreadsheets, half-documented processes that run whole companies. A lot of AI value sits exactly there, in the places the shiny demos never quite reach.
The hard parts
There were hard parts, and pretending there weren't would make the rest of this story dishonest. Imposter syndrome, starting Python at 30 next to people who'd been coding since they were twelve. Math catch-up — linear algebra, calculus, probability — on paper, at kitchen tables, late. Income drop. The brutal 2025/26 data job market that felt, for a stretch, like it simply didn't want me. Rejections that stacked up without explanation.
Some of that is universal. Some of it is just what it costs to change directions mid-career.
Some mornings the train goes further
Four years after that Tatra trail, I still take the same morning train out of Hamburg sometimes. The platform hasn't changed. The light on the Elbe hasn't changed. The coffee is the same coffee.
But the train ends somewhere different now. Mornings that used to end with spreadsheets opening end with a model training, or a new chapter of the thesis, or a small piece of code that makes a small thing possible that wasn't possible before. Same morning quiet. Different destination.
Some mornings the train carries you further than you booked for. It turns out that is a good thing.