Kaan Gümele

I build end-to-end data and AI systems at Mapin Data in Istanbul — from database architecture and cloud infrastructure to the back and front end of a dashboard and the LLM and agent layer that sits on top. On most of it I am the person framing the problem and making the architectural call. I am also an MSc researcher in computer engineering at Istanbul University-Cerrahpaşa.

My research concerns security in multi-agent LLM systems: when two agents converse freely in natural language, how much task-external information can they carry between them — and how can that capacity be reduced at design time? Before this I worked on deep learning for medical imaging, where I have four publications.

I teach Docker and NLP, do part-time AI consulting, and write technical posts. For a detailed record, see my CV.