Full-Stack Software Engineer · Computer Engineering (FSMVU, 2026) · Istanbul
I build web and mobile products end to end: backend, APIs, clients, store releases and CI/CD. I use Claude Code every day inside a structured process: persistent project memory, my own commands and skills, and sprints run by parallel sub-agents. Architecture, data and security decisions stay with me. Agent output is merged only after tests and review.
LinkedIn · Engineering case studies
Kelime Ezberle İngilizce is an English vocabulary app for Turkish speakers. I built it solo and shipped 8 App Store releases (1.0 → 1.4.0) in its first month. Google Play is in closed testing. Vanilla JS SPA · Capacitor 8 · PHP API · Firebase Auth + cloud backup · RevenueCat subscriptions · AdMob · Cloudflare Workers · 26 automated test suites (~1,500 checks) + native Android E2E. Case study
- BizSiz / Koçluk Sistemi (bizsizsoft.com): an ed-tech startup I co-founded and built the software for. It is a multi-tenant education and coaching platform with institution, teacher, coach, student and parent panels. It has tenant-level data isolation, 100+ migrations and 18 GitHub Actions workflows (static analysis, migration lint, daily schema-drift check, schema-gated production deploy with automatic rollback, anonymised staging refresh). A white-label Flutter app (Biledem Koçluk) is on the App Store.
- SEY Silver (seysilver.com): jewellery e-commerce with a PayTR iFrame payment integration (going live), variants, stock and a returns flow. 180 Playwright E2E tests run across 8 viewports plus Firefox and WebKit.
- İKRA Vakfı (ikravakfi.org), Pes Toptan (pestoptan.com.tr) and Biledem (biledem.com): production sites and admin panels (PHP 8 · MariaDB).
| Repository | What it is |
|---|---|
| claude-code-engineering-skills | How I work with AI: playbooks from real projects (sprint orchestration, blind second review, LLM evaluation, CI gates), session-memory commands, skills, hooks and tools |
| engineering-case-studies | Architecture write-ups: Kelime Ezberle, an LLM game master evaluation, BizSiz, SEY Silver and more |
| tea-house-tycoon-game | Tea House Tycoon, a 3D idle-tycoon mobile game (Google Play closed testing). TypeScript · React 19 · React Three Fiber · Capacitor · Vitest |
| watchsync-ai | Multi-channel inventory and listing platform, built by a two-person team. Laravel 11 · Next.js · FastAPI |
| soru-dunyasi | Gamified exam-prep platform. PHP · MySQL |
- Session memory protocol. A rule table in
CLAUDE.md, a line-capped state file that keeps the last 3 sessions and archives older ones, decision registers, and resume/finish commands. I use it in more than 10 projects, synced between two machines. - Sprints with sub-agents. Plan → design approval on a live preview → up to 3 parallel agents with written briefs → I merge → full test run → a changelog of every user-visible change.
- Measure, don't trust. AI-generated content gets a blind second review with an error threshold. LLM features get measured: 3 models over 55 scenarios (495 LLM-judged turns) before choosing a game-master model. See the LLM evaluation case study.
Stack: PHP 8 · Laravel · TypeScript · React / Next.js · Python / FastAPI · Flutter · Capacitor · Unity / C# · MySQL / MariaDB · Firebase · Cloudflare Workers · Playwright · Vitest · PHPUnit · pytest · GitHub Actions · Docker