Unified Models · Agents · AI for Health
I work on unified multimodal models and agents for health. My research connects medical image understanding and generation with evidence-grounded reasoning, and explores how evaluation can guide better data collection and model improvement.
Beyond healthcare, I collaborate on multimodal reasoning and scientific foundation models. I am currently exploring how agents and models can work together and improve through feedback.
- Unified understanding and generation — connecting image interpretation, synthesis, and multimodal reasoning within shared models.
- Medical agents and evidence-grounded reasoning — integrating visual and textual evidence to support medical tasks.
- Generative modeling and data synthesis — producing controllable medical images and studying their value for downstream learning.
- Evaluation and model improvement — identifying failures in robustness, calibration, and reasoning, and using them to guide improvement.
-
UniMedVL · ICML 2026 · First author, equal contribution
Unifying medical multimodal understanding and generation in a single model through an Observation–Knowledge–Analysis framework. -
MedQ-Engine · MICCAI 2026 · Co-first author
Turning failures in medical image quality assessment into targeted data collection, annotation, and iterative model improvement. -
RetinaLogos · MICCAI 2025 · Co-first author
Generating high-resolution retinal images from detailed descriptions of anatomy and pathology. -
Intern-S1 · Technical report · Co-author
Contributing to a scientific multimodal foundation model spanning general reasoning and specialized scientific tasks.
See my homepage for more work on medical model robustness, multimodal benchmarks, and scientific reasoning.
I completed my MRes at Imperial College London with Distinction and my BSc (Honours) in Data Science at the University of Sydney with the University Medal. My earlier training at the University of Melbourne was in Mathematics and Statistics, alongside a Diploma in Computing.
My research experience also includes work with the General Medical AI (GMAI) group at Shanghai AI Laboratory.
Homepage · Google Scholar · Hugging Face
I welcome conversations and collaborations on unified multimodal models, medical agents, and related problems in AI for Science.