PLAYER 01 · CMAP / APPLIED MATHEMATICS
How can AI learn from scientific images without losing the rules of the world?
I study generative and uncertainty-aware machine learning for polarimetric medical imaging. The public version is simple: make models expressive, physically coherent and honest about what they do not know.
Polarimetric imaging
Instead of treating an image as only colour and brightness, polarimetric imaging records how light interacts with a sample. That extra structure can reveal information that ordinary images miss.
INPUT: LIGHT · OUTPUT: STRUCTUREModels that imagine carefully
I explore VAEs, flow-based models and related generative methods to represent complex scientific images and create useful variations when labelled data are scarce.
LATENT SPACE / CONTROLLED VARIATIONPhysics-informed learning
A generated image can look convincing and still be impossible. My work studies how mathematical and physical constraints can be built into the representation, the decoder and the evaluation.
REALIZABLE OUTPUTS / CHECK THE CONSTRAINTSUncertainty and low-data medicine
When datasets are small, uncertainty is part of the result. I compare what models predict, where they become unstable and whether generated data actually help a downstream segmentation task.
CONFIDENCE IS NOT CERTAINTYWhat I can share here
The research direction, the mathematical questions and the ideas behind the models.
What stays private
Patient data, identifying information, unreleased experimental details and claims that still need independent validation.