Research

ClarifAI is an LLM-based system designed to collect structured feedback in hard-to-reach environments such as busy clinical settings, where traditional feedback mechanisms fail. By embedding AI-powered clarification dialogues into clinical workflows, ClarifAI enables continuous, contextual improvement of decision-support tools at the point of care.

Privacy-preserving Large language models for Acronym Clinical Inference and Disambiguation. A framework for inferring and disambiguating clinical acronyms at scale, entirely on-premises, improving EHR readability and downstream NLP task performance without exposing patient data.

A web application running on a local server that enables clinicians to monitor recovery data from post-operative patients between scheduled appointments. By surfacing structured trends and threshold alerts directly to the clinical team, MedWatch supports earlier intervention without requiring direct EHR integration, an important enabler in settings where institutional or governance barriers make full integration impractical.

A drug-drug interaction decision support project in collaboration with the University of Utah and Vanderbilt University. DDInteract focuses on presenting complex pharmacological interaction data in clinically actionable ways, improving prescriber awareness and reducing adverse drug events.

An AI-powered HCI evaluation system that deploys 10 specialist agents simultaneously, each an expert in a distinct evaluation framework, and returns colour-coded evaluation matrices, severity scores, and actionable recommendations in under 90 seconds. Built on Claude and designed to make expert-level UX evaluation fast, thorough, and accessible.

A user-centred mobile application delivering evidence-based clinical guidelines at the point of care, developed during my PhD in partnership with University Hospital North Midlands NHS Trust. Designed through iterative co-design with clinicians, resulting in 15 published usability recommendations across 5 papers.