Ilkcan Keles has spent much of his career asking how a system decides what to return, how that decision can be evaluated, and what changes when the system leaves the laboratory.
Today, he is a Senior Machine Learning Engineer at Booking.com, working in its GenAI Evaluation and Responsibility team. Before that, his route crossed database research, university teaching, search engineering, and the infrastructure around production machine learning.
From Ankara to Aalborg
Ilkcan completed undergraduate and master’s degrees in Computer Engineering at Middle East Technical University. His early work combined research with responsibility for real computing systems, including departmental servers and a high-performance computing cluster.
He later moved to Aalborg University, where he completed a PhD in Computer Science in 2018. The research examined spatial keyword querying: how a search system can rank results when both location and language matter.
The interesting part is not location data alone. His dissertation also dealt with ranking evaluation and the question of how user preferences can be represented. Those are recurring engineering problems. A system needs a way to choose, but its builders also need evidence that the choice is useful.
Research that moved across systems
Ilkcan stayed at Aalborg as a postdoctoral researcher and later an assistant professor. His publication record includes work on spatial data, semantic-web systems, and knowledge-graph querying.
His public career page shows a parallel interest in implementation. At Turkcell, he worked on indexing and ranking for a search-engine project. He later joined TomTom’s Search Product Unit, taking location and search from academic questions into another production setting.
It would be too tidy to claim that every move followed one plan. What the public record does show is a consistent willingness to work across the boundary between a research question and the software that has to answer it reliably.
Evaluation as an engineering habit
At Booking.com, Ilkcan’s submitted biography describes work on responsible development as well as the tooling and workflows around machine-learning systems. That combination matters. Evaluation is not a final score attached after implementation. It depends on what was measured, which cases were included, and how the result changes decisions in a working system.
That is an editorial reading of his path, not a personal claim from Ilkcan. It is grounded in a career that has repeatedly brought ranking, querying, infrastructure, and measurement into the same frame.
We are glad to welcome Ilkcan to the first LatentShift edition in İstanbul on 17 October. Until then, you can explore his research and career archive or connect with him on LinkedIn.