I am a Research Fellow in the Faculty of Epidemiology and Population Health, where I work on methodologies that combine Natural Language Processing (NLP), Large Language Models (LLMs), and Information Retrieval (IR) to extract actionable structured information from social media and other real-world data sources. The main objective of the research is to support the estimation of hidden mortality and case burden of humanitarian crises and neglected diseases.

Previously, I conducted research in the areas of multimorbidity and polypharmacy at the Nuffield Department of Primary Care Health Sciences, University of Oxford, where I worked as an ML/AI Scientist. During this time, I developed an agentic AI software system, AIMed++, to identify patients with dementia who are eligible for Structured Medication Reviews (SMRs) and optimise their medications. 

Prior to Oxford, I was a Postdoctoral Research Fellow at the UCL Institute of Health Informatics, where I worked on the NIHR-funded AIM-CISC project in collaboration with the University of Edinburgh. My research focused on developing NLP methods for extracting geriatric syndromes from clinical notes. This research resulted in a robust tokenisation methodology for encoder-only Transformer models, such as BERT, for clinical text processing, named K-Tokenizer, which has been open sourced.

I completed my PhD in Natural Language Processing and Biomedical Informatics at Birkbeck, University of London, in 2022. My PhD research focused on developing machine learning methods for extracting actionable health information from social media. I developed several supervised, semi-supervised, and weakly supervised algorithms for efficient health information extraction using Conditional Random Fields (CRFs) and neural network architectures, including Long Short-Term Memory (LSTM) networks and Transformers.