Deep phenotyping
Studying human physiology using multi-omics, imaging, wearables, and clinical measurements.
Hagai Rossman.
Assistant Professor of Personalized Medicine
MBZUAI / Chief Scientist, Pheno.AI
I develop AI models of human physiology from longitudinal health data.
I am part of the core co-founding team of the Human Phenotype Project, where I help guide the science and computational research.
Research areas
Foundation & generative models
Training models on physiological signals and multimodal data to predict health changes and model responses to interventions.
Health & research agents
Developing AI agents that use biomedical knowledge, patient data, and analysis tools to interpret health measurements and conduct research.
Selected publications
The Human Phenotype Project
A longitudinal cohort combining molecular, clinical, imaging, and wearable measurements to study health and disease.
GluFormer
A foundation model trained on continuous glucose data to predict metabolic traits and health outcomes.
CGMap
Glucose reference profiles in adults without diabetes and their associations with retinal and sleep measurements.
PulseOx-FM
A foundation model trained on overnight pulse oximetry to predict physiological traits and cardiometabolic risk.
HealthFormer
A generative model of multimodal health data for forecasting, risk prediction, and intervention simulation.
PhenoBench
A benchmark for comparing predictive models across clinical, molecular, imaging, and wearable data.
Background
I earned my PhD in Computer Science at the Weizmann Institute of Science with Eran Segal. I used nationwide electronic health records to predict childhood obesity and gestational diabetes.
During COVID-19, I led data-science work in Segal’s group on population symptom surveillance, hospital forecasting and strain, and vaccination impact.