I lead the technical side of a sovereign AI platform at IBM for Europe's regulated industries. After bedtime — I'm a twin mom — I build the things I needed myself and couldn't find: an app for exhausted parents, tools that do my own busywork, a world championship that deserved to leave the school gym, and education formats that reached thousands of kids. All tracked truthfully on this page: the live ones, the broken ones, the paused ones. My school research paper slept twelve years before waking up as that live game.
Everyone posts the wins. This page posts the rest as well: the paused, the broken, the quietly ended — and the ones where I finally admitted maybe this one isn't for me.
My ship log is a little embarrassing and, at the same time, the most impressive thing I have. Both are true at once. That's exactly why I show it: we are all trying to find the thing that fits us, and nobody ever found it by reading other people's highlight reels.
By day I'm a Tech Leader at IBM on a sovereign AI platform: AI infrastructure for Europe's regulated industries — public sector, financial services — that runs where the data has to stay. I lead the technical side of bringing that platform to life: cloud architecture, DevSecOps, deployments measured in weeks rather than months, and a team culture where no question stays unanswered for long — because speed is culture before it's tooling.
Part of that work is AI enablement: vibecoding sessions and first agentic use cases that take teams from AI-curious to shipping. The line that guides all of it — at IBM, at chilld, everywhere: building safe spaces for humans in an AI world.
If scaling AI across an enterprise is your topic, start with IBM Enterprise Advantage.
A sovereign AI platform by day and consumer-facing side projects by night — plus volunteer work and a seat on the works council — people assume one of them must be the hobby and the rest the real thing. They're the same job at different speeds, and each one is the reason the others work.
Four wheels, one transmission — click a wheel to hold it
Scale I could never reach alone. A sovereign AI platform touches millions of people's data across a continent, under rules that exist for good reasons. You don't get to practise that in an evening.
Constraints that make you better. Regulation, governance, stakeholders in five countries — the hardest teacher I know. Everything I understand about safety as an engineering discipline, I learned where the stakes were highest.
A seat at the table where the rules are made. I also sit on the works council: what a company may measure about the people who work for it, and what happens to someone when things go wrong.
Runway. It funds the evenings — no investor, no scholarship, no incubator. My Skip List has a whole entry about that.
And the slowness is a feature. Consultation, four opinions before a decision, a meeting to prepare the meeting — often maddening, and often right. Some problems dissolve if you let time pass instead of acting immediately, and a decision that survives six weeks of objections beats the one I made alone on a Tuesday night.
Hands on the tools. I can't credibly tell an enterprise team how AI changes engineering if the last thing I built myself was a slide. The vibecoding sessions work because I vibecoded something on Tuesday night and know exactly where it breaks.
My research lab. Every technology I actually understand, I understand because I used it on something real. There is no faster way to stay in the loop than to build with the thing.
The whole loop, at my own speed. Idea to live in weeks, no committee. That's where I learn what "fast" actually costs — and why I can be honest about pauses, because I've taken them.
It decides what gets built at all. chilld exists because of a 2 a.m. question I couldn't answer. The Pentomino championship is a family game before it is a product. My best ideas didn't come from a strategy day; they came from being needed.
It enforces focus better than any method. With twins there is no "I'll just work through the weekend". Bedtime is a hard boundary, and hard boundaries are why anything ships at all.
It is the reason the slow speed is fine. A four-week consultation is bearable when the important hours of your day were never at the office anyway.
Where thinking actually happens. I bake when I think and run half-marathons; the answers arrive on kilometre eight or somewhere in the dough — never while staring at the problem.
Honest feedback comes from friends, not from an audience. "Would I actually use that?" is worth more than ten people clapping — which is exactly why my Idea stage asks for doubts instead of applause.
The smallest wheel is the one people remove first. It is also the one whose absence you notice last and pay for longest. Bouldering, friends, a nerdy love of fluid mechanics: not the reward for the work, part of the machine.
Four wheels of different sizes, one transmission: the big one turns slowly and carries the load, the small one turns fastest and learns quickest — and they only move at all because they touch. Take one out and the whole train stops, which is the least romantic and most accurate description of balance I know.
The honest answer underneath both: I want the places where people spend their lives — a bank's infrastructure, a parenting app, a volunteer team — to be safe enough that people dare to think out loud in them. That's one job. It just happens to be done in a data centre on Tuesday and on a kitchen table on Wednesday.
And there's a message in the arrangement itself: you don't have to quit your job to build things. I once spent weeks hunting scholarships and incubators so I could go all in. Today I know I need both — and not because of the money. The night projects are what make me good during the day. Every technology I actually understand, I understand because I used it myself on something real, not because I read about it. There is no faster way to stay in the loop than to build with the thing. Day job and evenings aren't a compromise — they're one setup.
Every project here is tagged with where it honestly is — the live ones, the broken ones, the paused ones, the ones I ended myself: Idea Build Launch Scale Maintain — because the path between those stages is anything but a straight line.
A GUIDED TOUR — OR CLICK ANY STATION TO DIVE IN YOURSELF
An idea isn't a project yet — it's an invitation to a first lap. Don't polish it, don't pitch it. Find the smallest thing you could put in front of a real person this week.
Before: notebooks full of ideas and no way to test one without weeks of work — so most stayed notebooks. Now: an idea can be something clickable the same evening. The excuse is gone.
You plus two honest friends. The classic mistake: pitching the dream and fishing for applause. Ask for doubts instead — one "hm, would I actually use that?" is worth ten claps.
The video tool. I love editing — laying music under a cut until the right emotion lands is one of the most human parts of making things. AI isn’t at my quality bar for that yet; this tool is my attempt to close the gap, eventually fully agentic. That tension between love and tooling is exactly what a good idea feels like.
FastBrand is pure Idea-stage right now: an AI creative director that delivers a brand kit in minutes and sets up socials with templates — then hands over to Sendely. First lap not yet driven; that's allowed here.
chilld was born during parental leave — the same leave in which the book happened. Ideas don’t wait for calm phases; they show up mid-chaos.
One lap = build → show a real person → learn → adjust. Lean Startup calls it build–measure–learn; I call it Tuesday evening. A lap without a real person in it is just polishing.
Before: one lap cost six weekends of writing every line yourself — this is where most projects died, quietly. Now: AI writes the boilerplate, you make the decisions. Laps are so cheap you can afford to be wrong often — which is the whole trick.
Your first testers. The classic mistake: building alone too long and only showing the polished version. The ugly version teaches more — show it earlier than feels comfortable.
Sendely. I’m automating my own social media — my voice, database-first, so no task ever has to be redone. Building for yourself first keeps the feedback loop brutally honest.
chilld taught me the expensive lessons. I started with a co-founder and we split over direction — I had looked for one partly because the road alone felt “lonely” and “hard”. Shared suffering only helps if the other person takes suffering away instead of adding it. And I trusted the wrong people early; some of the money left with them.
Launch is a season, not a day — and it repeats on purpose. A waitlist is a launch. A founding circle is a launch. The App Store is a launch. You're in Launch the moment one stranger uses it without you watching.
Before: one big expensive launch — feedback arrived after the budget was spent. Now: launching is a habit. Feedback arrives while you can still act on it.
Your first strangers. The classic mistake: treating feedback as an attack — or leaving questions hanging. The first ten users decide whether your product feels alive; answer them in minutes.
Pentomino Duel is in Launch right now: live since early August, first real users — and honestly, it's a scaffold that breaks every other day. The bar is high: a championship that has run since 2004 deserves better than my current uptime. And the next laps are already scheduled — paywall, paid acquisition, an online world championship. That is what Launch looks like from the inside: shipping the next thing while the last one still wobbles.
chilld’s hardest lap isn’t willingness to pay. I’ve tested several formats and every single one found at least one paying customer. The hard part was admitting, each time, that the format itself didn’t work — for reasons that had nothing to do with the price — and ending it myself instead of waiting for it to die. Right now: founding testers, and the search for the format worth keeping. The honest blocker along the way was fear of loud outreach; overcoming fear is on my active list.
My Skip List before validation — thirteen things that felt like progress and validated nothing, from lawyers-before-customers to "I know better than the process". It earned its own section: the Skip List.
BYTE’s early mistake: not working closely with parents and experts from day one. And onboarding decides everything — one click too many, and people are gone.
The gate to Scale has a name — product-market fit: the market pulls, and no amount of hustle fakes the pull. You're in Scale when it grows while you sleep. And if it never comes, Maintain — small, useful, loved by a few — is a win too.
Before and now: still human-gated — the market decides. What changed: keeping several projects alive until one gets pulled is finally affordable for people with day jobs and twins.
A real team — at BYTE, 50+ volunteers. The classic mistake: the founder as bottleneck. If every decision passes through you, you've built a job, not an organization. Hand out roles, not tasks.
BYTE scaled with zero ads: teacher mailing lists, cold emails, social — that’s how 2,000+ students happened. In future apps I’ll test paid acquisition too, and share the numbers: what works, what doesn’t, mistakes included.
Leading 50+ volunteers taught me (straight from my BYTE team keynote): personal development is the best salary. Onboarding is a weeks-long process that ends in a personal bond, not a checklist. And you hand out creative latitude and goals — never task lists.
The audience lesson I’m still practicing: my instinct says keep it broad, reach everyone. The hard learning: one sharp target group, made genuinely happy, builds the better product. As a mom of three-year-olds I no longer know a newborn mom’s problems — personas beat assumptions.
Some things are finished. They run, they are useful, and growing them would only make them worse. Maintain is not the waiting room for Scale — it is a destination, and most lifecycle models pretend it doesn't exist.
Before: a tool for an audience of one was never worth building — the effort only paid off if many people used it. Now: a weekend of work can serve exactly one person for years. AI didn't just make products cheaper; it made personal software rational.
Often nobody but you. The classic mistake: feeling guilty about it, or scaling something out of pride that nobody asked for — and turning a tool you loved into a product you have to support.
FastBrand is headed straight here, on purpose. I'll build it, I'll use it, and I will not spend a single evening scaling it. An audience of one, forever — and that is a finished project, not an abandoned one.
Pausing is allowed. Abandoning is optional. Projects may sleep for years — mine do. My Pentomino paper slept twelve years and woke up as pentomino.club, opening a championship that lived in one school gym to anyone with a browser.
Before: coming back after two years meant a rewrite, so pause quietly became abandon. Now: waking a project costs a weekend. "Never touch it again" almost never has to happen.
A community that deserves one sentence. The classic mistake: silent ghosting. An honest "this sleeps for now" keeps trust; disappearing spends it.
The children's book tool is my favorite pause: I wrote and published the Bitsy books on Amazon — the writing was joy, the details were grind, and honestly, my kids and I don't love the current editions (please don't buy them yet). It sleeps until an agent can carry the annoying parts. That's not giving up; that's timing.
GI governance & Europe is asleep right now — said openly, on this very page, instead of ghosted. That’s the whole trick: a pause you announce stays a pause.
Two honest endings live here. The healthy one: it stops needing you. And the underrated one: this one isn't for me — the format worked, the market was there, and you still decided it wasn't the life you wanted. Stopping on purpose is a result, not a retreat. Hand it over in good shape, stay reachable, let them change what you loved. I did exactly that with the BYTE Challenge — and stayed as a mentor.
Before and now: handing over well was always about people, documentation and trust. What changed: AI helps with the documentation. The trust part stays yours.
Successors — and you, as mentor. The classic mistake: leaving without a handover, or never truly leaving: shadow leadership from the sidelines.
STEM GIrls worked because we solved people’s real problems with them — it recruited new people into the GI and built friendships that still hold. The best part: it found the person whose heart-project it became. Letting go was the success.
Gesellschaft 4.0 taught me two things: never underestimate conference logistics next to a day job — and nothing, truly nothing, replaces meeting people in person.
The point of the whole map: don't abandon the project. AI has made the laps radically cheaper — what used to die in the "someday" drawer now ships on a Tuesday night. This map is roughly 80% of my talks, and it's yours, free. If you want the other 20% live, bring me on a stage.
Build
Launch
Exit
An AI creative director: a solid brand kit in minutes — logo, colors, type, templates for every asset. The preview is live and free: answer nine questions and watch a brand appear, everything computed in your own browser. Destination: Maintain — I'll build it, use it, and never scale it. An audience of one is a finished product, not a failed one.
Launch
Your parenting village, finally. Anonymous peer support, real experts and AI-assisted voice notes — so a 2 a.m. worry gets a real answer instead of a doomscroll. Founder and CEO; testing with the first parents right now — message me if you want in :)
EVERY FORMAT I TESTED FOUND A PAYING CUSTOMER — AND I STOPPED IT MYSELF
Launch · live
A real world championship since 2004: 80+ students a year at a school in Saarlouis, under rules found nowhere else. I grew up in it and won it; my siblings hold titles too. My 2014 research paper about the game is now a web app anyone can enter — duel an AI on four levels, play hotseat, or challenge someone across the world.
LIVE SINCE AUG 2026 · WEEK ONE · NEXT LAPS: PAYWALL, PAID ACQUISITION, AN ONLINE WORLD CHAMPIONSHIP
Build
A social-content engine: speak a voice note, get a publish-ready post — drafting, swipe-to-approve, scheduling, analytics. I'm building it for my own channels first. If it keeps working this well, it won't stay private.
A video-creation pipeline that goes from script and raw material to a finished, branded cut in minutes instead of days — titles, captions, music, the lot. Built because I needed it; kept because it works. The icon is a question mark until the name earns its place.
I've done this the hard way already: wrote and published the Bitsy picture books on Amazon. Writing was pure joy; the details — layouts, uploads, endless revisions — turned it into grind. Honest verdict: my kids and I aren't in love with the current editions, so please don't buy them (yet). Restarting soon: with an agent doing the annoying parts, this might finally be as fun as it deserves to be.
Thirteen things that feel like progress before validation — and validate nothing. I did almost all of them, so this list exists. Skip them. Until real customers say otherwise, don't:
Download the long version (PDF) → PER ITEM: WHY I DID IT · WHAT IT COST · WHAT I'D DO NOW
Safe spaces aren't soft — they're the hardest infrastructure to build.
In 2025, I was robbed and assaulted in Berlin. That made concrete what had always been abstract: safety isn't a nice-to-have. It's everything — at work, at home, online. So I build systems where people feel genuinely protected: sovereign cloud architectures, learning platforms where no question feels "stupid", a parenting app that is anonymous before it is anything else.
chilld is anonymous before it is anything else: you can ask a question without an account, because the order in which we built it was the decision. Accounts came later, and only where experts genuinely needed a profile.
An answer isn't a ticket closed — it's someone stopping because your question mattered. Three answers, and the room has told you that asking was right. And it runs both ways: everyone who answers leaves warmer too, because helping is what belonging feels like. On the other side, the question fades, and with it the next one — and nobody there ever got to be the helper. That's what I'm really counting.
At the BYTE Challenge nobody had to answer me — 50+ volunteers, no salary, no hierarchy, no leverage. The only currencies that existed were meaning, ownership and being answered fast. Everything I know about leading without authority I learned there, not from an org chart.
AI can retell a joke. It can't feel the half-second of silence before the punchline lands, or sense that this room, today, needs a softer one — timing, reading the room and shared vulnerability are exactly what machines are worst at. Same for comfort: knowing when to say nothing at all. So I put small grins into serious products on purpose. If software can make a tired parent laugh once, it has earned its place.
The BYTE Challenge teaches computer science to school students, and its face is a small turquoise robot called Bitsy. He came on stage with me at German UPA 2023 and sat on the lectern through the whole keynote. A curriculum does not need a mascot. The children who stayed did.
Press the button — that is the whole argument
Questions, admissions and dissent are floors. Take the slab away and they don't get worse — they tilt and stop.
Onboarding at BYTE was deliberately slow: all digital, no prior knowledge required, everyone got told the story, the goals and the tools — and it took weeks, ending in a personal bond rather than a tool login. That was the slab under everything the volunteers later dared to do.
It doesn't stop the fall — it decides how high people jump
Safety isn't comfort. It's permission to try the thing that might not work.
STEM GIrls drew 367 sign-ups in its first year for the least comfortable topics there are: salary negotiation, public speaking, leaving your comfort zone, switching careers. Nobody signs up for that because the topic is easy. They signed up because the room was safe.
My project stages, applied to a person's relationship with their work
Nobody starts trusted. Trust is earned in laps, exactly like a product — and the same two stages that every project model forgets are the ones people are punished for: Pause and Exit. Name them, and parental leave stops being a gap in a CV and becomes a stage with a documented way back.
I've been in Pause more than once — a GI role sleeping right now, a book project waiting for better tools, parental leave in which I wrote a book anyway. And I've done the healthy Exit: I left the day-to-day of the BYTE Challenge and stayed as a mentor. Neither shows up as failure on this page, because neither is.
The famous 2×2, with talent instead of pressure on the second axis
The literature pairs safety with pressure. Pair it with talent instead and the quadrant everyone recognizes appears: brilliant people delivering while afraid, on a countdown to burnout. Every quadrant needs a different fix — and only one of them is a performance problem.
Leading 50+ volunteers at the BYTE Challenge, where nobody was paid: people who got roles and creative latitude thrived; people who got task lists went parked, then quiet, then gone.
Trust, as a number you can read off a clock
Every model above needs a questionnaire. This one needs a clock. Trust is visible in how long an unanswered question survives in a channel, and it decays in a predictable order: people ask freely, then carefully, then they guess, then they stop asking — and by then you've lost the information, not just the mood.
It's how a four-week deadline got met: not better tooling, but a team where no question stayed unanswered for long.
Four talks — practical, personal, never off a template. Each also works as a panel or a hands-on workshop, in English or German.
The honest project lifecycle in the AI era. Idea and Build chase each other in circles, launch happens three times, and scale may never come — none of that is failure. Told through my own projects: the school paper that slept twelve years before becoming pentomino.club and put a 2004 world championship online, and why AI changed the physics — laps that cost six weekends now cost an evening.
KEYNOTE · WORKSHOP · EN/DE
FOR FOUNDER PROGRAMS & INNOVATION TEAMS — YOU LEAVE WITH THE STAGE-TAGGING FRAMEWORK AND THE SKIP LIST
Psychological safety in tech teams. Grounded in a personal story I tell on stage — two minutes of context, then engineering discipline: people can only think, ask and build where they feel safe, so we treat safety like infrastructure. A metric, failure modes, and fixes for Monday — built on three maps I drew myself when the existing ones turned out to have no pause in them. Not an HR poster.
KEYNOTE · PANEL · EN/DE
FOR ENGINEERING ORGS & LEADERSHIP OFFSITES — YOU LEAVE WITH THE MINUTES-TO-ANSWER METRIC AND THREE FAILURE MODES TO FIX ON MONDAY
AI enablement that survives the workshop high. Most organizations don't have an AI knowledge problem — they have a first-rep problem. What actually happens in a vibecoding session, what changes about engineering judgment, and how a team gets from slideware to its first shipped agentic use case.
KEYNOTE · WORKSHOP · EN/DE
FOR ENTERPRISE TEAMS STARTING WITH AI — YOU LEAVE WITH YOUR TEAM'S FIRST SHIPPED AGENTIC ARTIFACT, NOT SLIDES
Parenthood, goal setting and returning on your own terms. I published a book during parental leave with twins — and the point of that story is not the hustle. Comebacks aren't about doing more: they're about honest stages, systems that forgive pauses, and goals that survive a newborn phase. For returners and for the employers who want to keep them.
KEYNOTE · PANEL · EN/DE
FOR RETURNER PROGRAMS & EMPLOYERS — YOU LEAVE WITH GOAL SYSTEMS THAT SURVIVE A NEWBORN PHASE
Booking: write me with "Speaking request". Also on my stage list: digital education & women in tech — lessons from BYTE and STEM GIrls. And I mentor new parents and founders through the transition back into work — reach out if that's you.
I'm a computer scientist (M.Sc., TU Berlin) living in Berlin and a Tech Leader at IBM. Everything else on this page happens around that: founding chilld, co-founding the BYTE Challenge, years of volunteer work in the German Informatics Society, writing books, and giving keynotes.
The through-line is speed with honesty. I'd rather put a real, small thing in front of real people this week than a perfect thing in front of nobody next year. Every product on this page started as a personal itch — a parenting question at 2 a.m., a family board game, my own social-media backlog — and got vibecoded into the world before the doubt could catch up. AI didn't change what I build; it removed the excuse not to.
If you need a label for all of this: a human tech leader in an AI world. But I'd rather the receipts say it — the ship log above, the minutes-to-answer metric, the products that ask one kind question instead of twelve cold ones. Same instinct whether I'm architecting AI infrastructure or answering a scared new parent at 2 a.m.
Off the keyboard: twin mom. I boulder, run half-marathons, bake when I think, and have a nerdy love for fluid mechanics. Born in Saarlouis, at home in Berlin. TEDx is on the list.
Unpaid work is where I learned to lead. At the BYTE Challenge nobody had to do anything I asked — 50+ volunteers, no salary, no hierarchy, no leverage. You find out quickly that the only currencies are meaning, ownership and being answered fast. Every leadership habit I have came from there, not from an org chart. The same is true in the Gesellschaft für Informatik: you convince, or nothing happens.
Not everything I build is software. Since 2019 I've been building formats inside the Gesellschaft für Informatik and beyond that pull more people into computer science — kids, women, non-binary people, career changers. Same lifecycle, same rule: don't abandon the project.
Co-founded in 2019 with Stefan Hildebrand: a free digital education platform with 100+ courses on computer science, digital skills and careers — fronted by Bitsy the robot. 50+ volunteers, 2,000+ students, €100k+ raised. I've since left day-to-day operations and stay close as a mentor — the healthiest exit a project can have.
The workshop series I built during my time on the GI board: role models, salary negotiation, public speaking, career switching — for women and non-binary people on their way into tech. 367 sign-ups across nine workshops in the first year alone.
The first-ever conference of the Junge GI, which I helped initiate: four days in Tutzing with the Academy for Political Education, bringing young computer scientists into dialogue with politics, business and science about generation-fair digitalization.
Spokesperson 2020–2022 and elected to the GI board (Präsidium) alongside. Founded the Berlin-Brandenburg university group and ran 33+ digital events with partners from IBM and Microsoft to McKinsey and GitHub. All handed over to the next generation — which was the point.
Structural change in the GI — statutes, tools, ways of working — through the AK Struktur I led from 2023, plus the women-in-computer-science group and its award, and European digital-skills policy at CEPIS. Currently on pause: the products above get the night hours right now, and pausing is allowed.
A movement pushing for real digital education across Germany — coding basics, digital sovereignty and media literacy for every student. I'm one of its speakers, arguing that enthusiasm for digital topics has to be sparked much earlier.
Children's stories in which Bitsy the robot solves everyday adventures — baking cookies, finding a lost cat, building a snowman — with ideas from computer science. Every book ends with a note for parents explaining the concept behind the story. Full honesty, ship-log style: the current editions aren't at my bar — they're getting rebuilt, and I'd rather you wait for the new ones.
The German career guide to the IT industry, written with Lisa Ihde and Junus Ergin — and published during my parental leave with twins. Insider knowledge and shortcuts for career changers and students — and a clean-up of 24 stubborn myths about working in tech. (In German.)
Twenty-four chapters, twenty-two listed here — the two missing ones are a reading incentive. Written with Lisa Ihde and Junus Ergin, Haufe 2025.
Booking a keynote, bringing vibecoding to your team, joining chilld's founding circle — or just want to compare Pentomino openings? Write me.