About

I build enterprise products and stay close enough to the code to know whether the plan survives contact with reality.

My route here was not the usual one. I studied American Studies, started out coordinating education programmes, and moved into ERP training — teaching several hundred professionals how enterprise systems actually behave. That turned out to be the most useful product education I could have had. If you want to understand where software fails people, watch someone try to use it while you explain it.

Compliance taught me systems thinking

I then spent time as a Quality Systems Manager, running certification and audit programmes — ISO 27001, TISAX, SPICE, and others — across an ERP vendor. It is not a background most product people have, and it is more relevant than it sounds. Audits force you to prove that a process does what you claim it does, with evidence. That discipline maps directly onto shipping software in regulated industries, where traceability is not paperwork but the thing that lets you ship at all.

The most satisfying work from that period was not the certificates. It was integrating business process management directly into the ERP platform and cutting approval times by ninety percent — a process problem solved with software rather than a policy document.

Consulting, then building

At DefineX I have worked across insurance technology and automotive aftersales — requirements, test strategy, UAT, and multi-market rollouts across fifteen-plus countries. Consulting sharpens a specific skill: walking into a domain you do not own, finding the real constraint quickly, and being useful before you are comfortable.

In early 2026 that turned into something different. I had the training-management domain in my head from years earlier, and a growing conviction that an AI-first delivery approach could compress a build that would normally take a year. So I built a mockup on my own and pitched it. It became a funded product with a team, and I became its product manager. Three months from the first commit it was in production, replacing the legacy system. It now serves more than 450 users.

On AI-native delivery

I use LLMs across the full lifecycle — specification, requirements, code review, test design, release documentation. What I have learned is that the interesting problem is not generation. It is validation.

A model will produce something plausible on demand. Plausible is not the same as correct, and the gap between them is where projects quietly go wrong. So hallucination risk gets treated as a first-class engineering concern: iterative validation, explicit guidelining each sprint, and test gates that do not care how confident the output sounded. That is the part of the methodology I would defend.

The other eleven years

Since 2015 I have led international advocacy for HAÖDER, the Turkish patient association for Hereditary Angioedema — a rare and potentially life-threatening condition. I represent Turkey at international rare-disease forums, and I built the association’s digital operating model: documentation, compliance workflows, and tooling that a volunteer-run organisation can actually sustain.

It is the work I am most careful about. Building something people depend on is a different discipline from building something people evaluate, and I would not have learned the difference anywhere else.

Now

Based in İstanbul. Interested in product roles where the product is genuinely technical, in AI-native delivery done with rigour rather than enthusiasm, and in enterprise modernisation where the legacy system is the actual problem.