Mark.
AI Engineer. Systems Architect. I build the parts of AI products that have to work at 3 AM when nobody's watching.
I'm a software engineer who ended up deep in LLM infrastructure — not by accident, but because the problem of making language models behave reliably at scale turned out to be genuinely hard, and I like hard problems. For the past few years, that's meant building multi-tenant chatbot platforms, retrieval pipelines, agentic loops, and the observability layer that ties it all together.
Before AI infrastructure, I was building product search systems in Ruby on Rails, debugging ECS timeout errors at 2 AM, and figuring out why jieba was tokenizing traditional Chinese differently from simplified. The common thread: systems where correctness isn't optional and cost is always a constraint.
The best AI infrastructure is invisible — it works so reliably that the product team stops thinking about it and focuses on what they're actually building.
I moved to Abu Dhabi to join G42 / Presight AI — one of the most serious AI bets in the region. The Arabic AI market is genuinely underdeveloped relative to the opportunity, and being inside one of the organizations trying to change that felt like the right place to be.
I write about what I actually ship — the writing section is a direct log of production problems, architecture decisions, and the occasional market analysis. No conference-talk summaries.