Darmstadt, Hesse, Germany
Most IT projects don't fail on the technology. They fail because no one ever pinned down what was really needed, and because complexity quietly piles up where nobody's looking. If you've watched an AI or digitalization push look clean in a boardroom and then stall the moment it meets real systems and real people, you know the feeling. That gap is where I work. I sit between the business and the people who actually use the systems, turn their tangled, real-world needs into a plan everyone can follow, and build IT that stays simple enough to help. Simple in the sense that it gives people room to do their jobs, and gives the company numbers it can point to. I've done this in places that look nothing alike. I've taken waste out of automotive development, and I've stood up business-critical applications in rail infrastructure: the dispatching and control tools the operators rely on. If they go down, the trains keep running, but the people steering the network lose their support and fall back to slower, manual workarounds. So "mostly reliable" was never the bar. Today I run a department of more than 30 people, and a lot of that job is making sure a strategy that sounded clean in a meeting still holds up in the engine room: that it scales, runs, and still makes sense once real systems and real people are involved. The methods that work today won't carry us through what comes next, so I don't stop there. Outside the day-to-day I keep a kind of tech lab, working hands-on with large language models, MCP servers, and autonomous agents. No slideware, no hype cycle. I build with these tools myself, because that's the only way to see where they genuinely earn their place and where they don't. What I'm working toward is easy to say: not just keeping processes running, but rebuilding them around AI and digitalization. I don't think the goal is to replace what people know. It's to make their expertise count for more. If that's the problem you're wrestling with, I'm glad to talk.
The teams who plan timetables, sell capacity and run daily operations at DB InfraGo can't afford to steer blind. When decisions hinge on data that's scattered, slow or hard to trust, good people end up reacting instead of leading and in a network this size, every hour lost in a performance review or a dispatch call is felt on the track. That's the problem I exist to solve. I lead the IT function that gives these teams one reliable place to see what's happening and decide what to do next. With a team of 35 (17 internal, 18 partners), I'm accountable for a portfolio of 35 business-critical applications and the enterprise data warehouse that DB InfraGo's timetable, sales and operations functions run on keeping them stable, fast, and evolving with the business. How I lead: build platforms the business can actually steer with, not dashboards nobody opens; bring internal and partner teams together around the outcome, not the org chart; and make every improvement stick through continuous improvement, not one-off projects. The result I'm after is simple: the people running Germany's rail infrastructure decide faster, with more confidence, and spend their energy moving the network forward not chasing the truth inside their own systems.
see current position
see current position
In a business like rail, operational leaders live or die by two questions: where is capacity about to choke, and where are we losing punctuality? When the answers sit in scattered reports or worse, in a queue waiting for the BI team managers steer blind and react too late. That's the gap I closed. As a Business Intelligence expert, agile project lead and Product Owner, I gave the business the numbers and the self-service tools to answer those questions itself, in real time. What I delivered: - New operational steering KPIs including capacity bottleneck and punctuality loss so operations leaders could see trouble forming and act before it hit the timetable. - A Tableau-based self-service platform with proper rights management, so business units answered their own data questions instead of waiting in a reporting queue. - Strategic advisory to those teams on automating their reporting turning hours of manual report-building into a few clicks. I led the analytics team and owned the division-wide Tableau Server platform end to end running it agile as Product Owner, professionalising it with governance a
Every wind tunnel session is expensive, and it can only tell you so much: whether the shape in front of you is a little better or a little worse. It can't tell you whether a fundamentally better shape was sitting just outside the few you tried. For a vehicle program chasing every count of drag, that's the difference between a good car and the best one the design could have been. That's the gap I closed for GM's vehicle programs at Adam Opel AG. As a CAE aerodynamics development engineer, I gave the aero and design teams a way to explore the whole design space in simulation — and find the genuinely best shape, not just a local improvement. What I built: - A method to parametrise the vehicle shape, so hundreds of design variants spoilers, underbody, and more could be explored in simulation instead of one at a time in the tunnel. - Predictive models that characterised how the system behaved, turning expensive trial-and-error into a fast, targeted search. - Scripted automation of the process steps, which not only sped the work up but made coordination with the design department far simpler. The payoff was the part wind-tunnel-only testing can't reach: instead of settling for the local optimum nearest the starting shape, the teams could find the global optimum and exploit a vehicle's full aerodynamic potential. The result: a faster, leaner aerodynamic development process and better cars, because the teams were no longer guessing whether a better shape was still out there.
Before a finite-volume simulation can run, someone has to build the grid. Done by hand, structured grid generation is slow, fiddly and easy to get wrong, and every hour spent on it is an hour not spent on the engineering question that actually matters. My doctoral research set out to take that burden off people: I worked on automating the generation of structured grids for finite-volume methods, so the setup that used to eat days could be produced reliably and fast. The work rested on a solid mechanical engineering education with a focus on applied mathematics and numerical methods, the foundation that let me turn a hard meshing problem into a repeatable, automated one. And it grew out of a conviction that has run through my career ever since: that smart automation should remove the manual bottlenecks holding capable people back. The part I valued most was the people. I supervised 10 student theses at Bachelor and Master level, bringing young talent into real research, guiding their work toward results that mattered, and folding what they discovered back into the science. They moved faster because they had someone in their corner, and the research moved faster because of them.