Theoretical physics · Machine learning · Science communication
Emre Onur Kahya
Professor of Physics · Istanbul Technical University
I study quantum fields, gravitation and cosmology — and, increasingly, what learning machines compute on the inside.
IIFields & the early universe
Quantum fields on a curved, expanding stage
How does a quantum field behave when spacetime itself is dynamical? I compute quantum — including quantum-gravitational — corrections to fields during inflation, and work on observational tests that can tell gravitational theories apart.
IIILearning from physical data
When measurements become representations
Machine learning helps physics when the data, the baseline and the evaluation reflect the scientific question — from classifying multiband light curves to identifying electrons in ATLAS data and forecasting satellite conjunction risk.
IVInside the model
From what a model predicts to how it computes
My current research direction is mechanistic interpretability: which internal representations a model forms, and which of them it actually uses. Interventions such as ablation and activation patching turn pictures into testable causal hypotheses — when baselines and controls are taken seriously.
Schematic illustration — no real model is analysed here. This is an ongoing research direction, not an established result.
Mechanistic interpretability & reliable AI The MYZ 310E research project
VPhysics, out loud
Waves that reach people
Since 2021 I have talked about physics, the history of science and artificial intelligence on Teke Tek Bilim — in long conversations with Fatih Altaylı, Emrah Safa Gürkan and fellow scientists, and in my own series, Bilim Ekstra.
- Programmes
- 56
- Views on YouTube & Instagram
- 10.3M+
- Bilim Ekstra episodes
- 49
Teke Tek Bilim programmes with Emre Onur Kahya. Views on YouTube — these programmes with their short clips, the personal channel and other channels — as of 19 October 2026; Instagram views of the clips are added by hand (19 October 2026).
Most watched: Yapay zekanın fişi çekilebilir mi? — a short clip with Fatih Altaylı, 1.2M views. (opens in a new tab)
All programmes & lessons My YouTube channel (opens in a new tab)
Teaching · Fall 2026
MYZ 310E — Machine Learning in Physics
Six weeks of intensive foundations — statistical learning, regression and classification, neural networks and representations — followed by a research project in mechanistic interpretability, two presentations and a final paper.
Assessment
- Midterm examination
- 30 points
- Project presentation I* (provisional split)
- 15 points
- Project presentation II* (provisional split)
- 15 points
- Final project paper
- 40 points
- Total
- 100 points
Final-paper eligibility: at least 12/30 (40% of the midterm itself) and completion of both project presentations.
Fourteen weeks indicative plan
- Foundations
- Midterm & scoping
- Project presentations
- Project clinics
- Paper development
Indicative schedule — Foundations: weeks 1, 2, 3, 4, 5, 6; Midterm & scoping: week 7; Project presentations: weeks 8, 9, 12, 13; Project clinics: weeks 10, 11; Paper development: week 14.
Selected work
Ideas in print
2025
2024
2021
2018
2017
2008
About
A physicist, a teacher, a communicator
I study quantum fields, gravitation and cosmology, and develop machine-learning approaches to physical data. My current research direction asks how the internal mechanisms of learned models can be understood and tested. I also talk about physics and artificial intelligence for a broad Turkish-speaking audience.
Recognitions include the Turkish Academy of Sciences Outstanding Young Scientist Award (GEBİP, 2015), the TÜBİTAK Science Incentive Award (2016) and the Mustafa Parlar Research Incentive Award (2017).