I’m Osman. Senior product engineer based in Bonn, Germany. For the past seven years I’ve worked at Studitemps as the engineer bridging business and engineering — leading architectural consolidation, shipping AI features in production, and owning automation work that has moved multi-million-EUR revenue. On the side, I’m building ClearRank.io, an AI visibility platform for the next shape of search. Originally from Boani and Herzegovina. Father of two. Remote for most of the past decade.
What I’m doing now
Three threads run through the current work at Studitemps:
- Architectural consolidation. Leading a move from distributed microservices with event sourcing into a modular monolith with clearer bounded contexts. Phasing out the event sourcing layer, unifying databases, strengthening domain boundaries.
- AI in production. Multi-step agent pipelines, retrieval-augmented context, model fine-tuning, and tool-use patterns that have to hold up under real customer load. The work is less about picking models and more about making inference predictable at scale.
- Operational automation. Make.com and n8n workflows connecting engineering to business operations without dedicated development cycles. Real impact per hour of work is higher here than anywhere else.
ClearRank — building outside employment
ClearRank tracks how ChatGPT, Perplexity, and Google’s AI Overviews cite and surface brand domains. The bet: AI-generated answers are about to replace a meaningful share of what used to be Google’s #1 result, and nobody has solid measurement for that yet. This is my attempt at building it.
I handle every part of it — the engineering, the product direction, the landing page, SEO, go-to-market, content. Running a product solo teaches things employment doesn’t: sitting with real users when the signup flow is broken, deciding what to ship and what to cut, watching how far a feature actually travels before someone writes back.
How I think about software
- DDD reduces miscommunication, not complexity. The biggest wins in the systems I’ve worked on come from business and engineering speaking the same domain language. Aggregates and tactical patterns are secondary.
- Architecture earns its complexity. Most distributed systems I’ve seen — including the one I’m currently consolidating — were distributed because it looked right on the whiteboard, not because the coupling cost exceeded the distribution cost. Event sourcing especially should come with a warning label.
- AI is a production concern, not a feature. The hard parts of shipping LLM-powered features aren’t model selection or prompt wording — they’re determinism, auditability, and failure boundaries. I wrote more about that in Déjà Vu: From YAML to AI Agents.
- Shipping frequency is a diagnostic, not just a discipline. Every team I’ve worked on that couldn’t deploy daily had already accumulated too much risk to deploy safely.
Background
Career started in Rails (2014), moved through Node, Python, landed on Elixir/OTP. Backend systems for metals trading, parking infrastructure, vendor integrations, and student employment. In every role I’ve been closest to the business problem — full-stack capable, but most useful where domain complexity and data model meet.
Get in touch
LinkedIn is the best way to reach me.