Research Assistant
CurrentAt the University of Pennsylvania, I developed and refined a Variational Autoencoder (VAE) to reconstruct visual fields, improving both runtime and functionality. I extended the VAE to predict future visual fields, enhancing its clinical applicability. Additionally, I explored the performance of diffusion models in comparison to VAEs for glaucoma progression, focusing on efficiency and accuracy. My work also involved archetypal analysis of visual field data to classify progression patterns and improve model performance. To further optimize results, I integrated CNN structures into the VAE and collaborated closely with PhD students and professors to align computational models with clinical diagnostics.