The version of this story that gets told is usually linear: study → job → startup → success. The version I lived was not linear. Here's the map, as accurately as I can draw it.
The thesis years (2014–2016)
I was doing an MSc in software engineering with a focus on automated testing for mobile apps that generate sensory data. Not AI, exactly - though ML was involved. The thesis got rewritten three times because the problem kept reshaping itself. I didn't think of this as useful training at the time. It was the first time I spent extended periods sitting with a problem that didn't have a clean solution.
I built a text classifier during this period. 94% accuracy. Mostly useless in practice. My first real lesson in the gap between a benchmark and a product.
The startup years in Turkey (2016–2017)
I co-founded Trendoline while finishing my MSc. Social app. Solo full-stack: Xamarin for iOS and Android, plus the backend on Azure. We won a competition. We got cloud credits. We built architecture we later regretted.
The CTO title came early - I was 26. The gap between writing code and leading technical direction is wider than it looks. I learned this by falling into the gap.
San Francisco (2018–2019)
I went to San Francisco to work on Zillion Pitches. This was the decision that shaped everything after. Not because of what I built there - though building real NLP in production before it was mainstream mattered - but because of where I was. The ambient intelligence density of San Francisco in 2018 was unlike anything I'd experienced. Conversations at events, between meetings, at dinner. GPT-2 dropped while I was there. I watched it happen in real time with people who understood what it meant.
I learned that proximity to the frontier matters, not just for what you know but for how you think.
Draper University (2018–2020)
Becoming a Program Manager at Draper while continuing to advise and build was not a planned move. I was asked to help with a cohort and it became a two-year role.
What I got from training 1,000+ founders across 15 cohorts wasn't business lessons. It was pattern recognition. I watched a lot of people try to build things. I watched what separated the ones who shipped from the ones who didn't. The Hero Mindset work Tim Draper did with the cohorts turned out to be a surprisingly reliable predictor. People who could maintain conviction through the friction of building tended to get further than people who couldn't.
COVID ended that chapter. We moved the entire program online in three weeks. Harder than it sounds.
Countercheck (2021–2024)
The work I'm most proud of technically. We validated the problem before writing code. We raised the right amount to get to market. We built a system that ran YOLO at scale and got to 99.6% accuracy in production at a logistics facility, not in a notebook.
The patent came from making the architecture work reliably, not from inventing a new model. The clients - LVMH, Lego, Moncler - came from solving a real problem with demonstrable results.
Three years later: acquired. The chapter closed.
Wasteer (2024–present)
Joined as Head of Engineering. Different kind of challenge: the company had technical debt and organisational debt and a product that needed to grow. I became CTO within thirteen months.
The things I'd built and broken before were all relevant. The architecture decisions at Countercheck. The startup company building at Trendoline. The operator thinking from Draper. The NLP foundation from Zillion Pitches. None of it was wasted.
The pattern I'd point at
The non-linear parts of the path - Draper, the thesis rewrites, the pivot from Turkey to SF - look like detours. They weren't. They were the parts that made the subsequent work possible.
If I were advising someone earlier in a similar path: don't optimise only for the direct route. The skills you build in the unexpected places tend to be the ones that compound.
With gusto, Fatih.