GPT model trained from scratch on EHR codes of 57 million individuals for universal risk prediction and trial emulation.
section {
background: white;
color: black;
border-radius: 1em;
padding: 1em;
left: 50% }
#inner {
display: inline-block;
display: flex;
align-items: center;
justify-content: center }
Twitter thread summarising key results
Representation learning for EHR data, exploring graph structures and semantic embeddings applied to national-scale datasets
Phenotyping at scale across diverse biobank cohorts to power genomic & proteomic analyses for target identification, drug discovery, and precision medicine.
section {
background: white;
color: black;
border-radius: 1em;
padding: 1em;
left: 50% }
#inner {
display: inline-block;
display: flex;
align-items: center;
justify-content: center }