Most AI writing stops at the API call. Most mobile writing stops at the screen. I work in between: native iOS and Android in front, models and pipelines behind. Everything here is something I measured, built, or broke in production.
on-device inference · quantization · Core ML · MLX · LiteRT · hybrid routing
LLM orchestration · verifier loops · prompt caching · evals · cost engineering
Engineering
- Tokens per second is two numbers
- Mobil Geliştirme AI ile Bitmiyor, Yeni Bir Forma BürünüyorTR
- Ya Daha İyisi Varsa? Overengineering ve FOBOTR
- Bir Üründe Refactor Kararı Nasıl Verilir?TR
- Know-How: Kaos SenaryolarıTR
Leadership & Teams