Scaling Biological Representation Learning and Experimental Design

I gave an invited talk titled ‘Scaling Biological Representation Learning and Experimental Design’ as part of the Maths Life Sciences seminar series at the University of Manchester. I discussed scaling masked-autoencoders on microscopy images, recovering known biological structure from representations, and outline mechanistic interpretability techniques and experimental-design strategies for large models.