I am a Senior ML Researcher at Samsung Research UK. My research interests lie in the adaptation of large language models (LLMs) within resource-constrained environments such as smartphones. Particularly, I am interested to learn more about:

  • parameter-efficient fine-tuning for efficient adaptation of LLMs to downstream tasks (see our works on hypernetworks [EMNLP’24, ICASSP’25],
  • decentralized training of models including LLMs (e.g., federated learning) for privacy-preservation,
  • model merging to combine multiple LLMs or PEFT parameters for efficient deployment,
  • continual learning to train models progressively over time,
  • model compression techniques (e.g., quantization, pruning, and knowledge distillation) to reduce the footprint of LLMs.

Previously, I was as a Postdoctoral Research Assistant in the Computational Health Informatics group at the University of Oxford, led by Prof. David A. Clifton, where I worked on developing generative models for healthcare, particularly for disease progression modeling and synthetically generating electronic health records (EHRs).

I obtained my PhD from the School of Informatics at the University of Edinburgh under the supervision of Prof. Chris Williams. During my studies, I was based in the Alan Turing Institute where I took part in the Artificial Intelligence for Data Analytics project. As a part of this project, my thesis proposes probabilistic data type inference methods for tabular data. Prior to my PhD, I was a Research Assistant at the at Department of Computer Engineering of Bogazici University, working with Taylan Cemgil.