Research Interests
AI for Scientific Discovery
Physics foundation models, automating theory generation, interpretable physics models.
Theoretical & Computational Physics
Foundations of physics, quantum algorithms, relativity
Deep Learning & Symbolic AI
Transformer architectures, representation learning, and symbolic regression for interpretable models
Physics-Informed Machine Learning
Incorporating physical laws and symmetries into neural network architectures
Research Experience
My interdisciplinary background has shaped my research interests at the intersection of physics and machine learning:
I graduated from Boğaziçi University with bachelor’s degrees in both Physics and Computer Engineering, earning a 3.53/4.00 GPA with high honors.
For my bachelor’s thesis, supervised by Asst. Prof. İnci Meliha Baytaş and Asst. Prof. Arkadaş Özakin, I developed an improved Constrained Lagrangian Neural Network (CLNN) that learns physical constraints directly from data.
I am attending the EuCAIF Conference 2026 from 24–28 August, where I gave two talks and presented posters on stabilized Lagrangian Neural Networks and simultaneously learned constraints. View the talks and presentations.
Previously, I worked at Radboud University’s High Energy Physics group, where I developed transformer-based symbolic regression methods for interpretable modeling of gravitational wave signals and data-driven discovery of physical laws.
Before that, I interned at Forschungszentrum Jülich's PGI-8 Institute of Quantum Control, investigating mutual information as a measure of Neural Quantum State learnability using transformer architectures and information-theoretic methods for quantum many-body systems.
Earlier, I explored Lagrangian Neural Networks (LNNs) and symbolic regression under the guidance of Asst. Prof. Arkadaş Özakin at Boğaziçi University, showing how machine learning can infer theoretical quantities like the Lagrangian from data. I later presented these findings in a seminar I organized at my university.