About

I'm Serkan. Software Architect in Istanbul, ~15 years in. This site is the notebook where I learn LLM systems in public through open-model evals, small-model data pipelines, post-training experiments, and reproducible reports with explicit claim boundaries. I studied computational biology and bioinformatics, which is why some of the work here wanders into bioinformatics and research-use data tooling.

Short bio

Serkan Altuntaş is a software architect in Istanbul with a computational biology and bioinformatics background. He writes about LLM systems, software architecture, and build-in-public projects at serkan.ai.

What I do now

By day I lead the Platform and SaaS Substream at Togg-Trugo, where I work on Turkey's EV charging network — OCPP, high-availability backend services, the usual distributed-systems problems. By night and weekend I read papers, rebuild small models from scratch, and try to understand what I'm running.

The LLM work is deliberately evidence-first rather than authority-first: small eval suites, frozen baselines, data cards, failure audits, and public notes about what the results can and cannot claim.

I am not presenting the biomedical projects here as clinical work; they are research-use attempts to make public evidence more inspectable.

Why this site exists

Writing slows me down enough to catch the places I'm confused. Publishing forces the writing to actually make sense. So most posts here are me working through something I don't fully get yet — not a tutorial from someone who does.

Stack I reach for

Python · Go · Rust · Swift · PyTorch · Postgres · AWS / GCP.

Elsewhere

Based in Istanbul, Turkey.