Fernando
Kavinsky

Engineer and M.Sc. researcher in Artificial Intelligence at the Federal University of Rio Grande do Sul (UFRGS), working with the ExpLAIn laboratory. My research focuses on how populations of neural networks move, reorganize and converge through high-dimensional weight spaces.

My background crosses software engineering, machine learning, distributed systems and aeronautical sciences. I am especially interested in methods that make complex learning systems observable, interpretable and reproducible — particularly where reliability matters.

Understanding how neural populations evolve through weight space.

The central problem: evolutionary algorithms produce populations rather than a single optimization trajectory. I investigate how dimensionality reduction, temporal alignment, optimal transport and vector-field representations can expose the redistribution of these populations over generations.

Aviation is both a personal pursuit and a research domain.

I study Aeronautical Sciences and am pursuing Private Pilot training. The technical intersection is particularly interesting: autonomous systems, trajectory optimization, anomaly detection and interpretable decision support for safety-critical environments.