Leonor Fernandes

Trained and optimized ML models (Logistic Regression, XGBoost, Decision Trees) for diabetic retinopathy detection. Designed deep learning pipelines using PyTorch and HuggingFace to analyze retinal fundus images. Researched and implemented autoencoders, VAEs, and GANs to generate synthetic images disentangled from patient attributes. Researched and implemented fairness metrics and visualization techniques to evaluate model bias, establishing a framework to detect and mitigate unfair models.

Built and maintained scalable Python data pipelines for processing multi-gigabyte neural datasets. Applied signal cleaning, feature extraction, and dimensionality reduction to improve decoding accuracy. Developed Kalman filter-based real-time decoding simulations to assess neural activity patterns. Applied manifold alignment to enhance decoding stability across sessions and subjects, validated in simulations.