Leveraging large language models to unlock and synthesize the expanding volume of medical research (MedLitGrasp)

Medical research is expanding faster than clinicians and researchers can systematically keep up with. This project leverages large language models to identify relevant evidence, extract key study information, and produce transparent, citation-grounded syntheses tailored to concrete clinical questions.

Key data

Description

The volume of medical publications is rapidly increasing, making it harder to retrieve and synthesise up-to-date evidence for practice and research. The project aims to use large language models to support literature search, information extraction, and evidence synthesis in a transparent, reproducible workflow.

We will build a pipeline for ingesting and curating literature collections, automatically prioritising relevant studies, extracting structured elements (e.g., population, intervention, outcomes), and generating citation-grounded summaries with quality and plausibility checks. The approach will be evaluated on defined use cases and iteratively refined based on expert feedback and measured performance.