Felipe Parodi

Research

My research connects three questions: how brains compute social meaning, how to build tools that make natural behavior measurable, and what we can actually tell about what large models are doing inside.

I study the neural and behavioral basis of social intelligence in primates — both human and non-human — combining wireless neural recordings, quantitative behavioral analysis, and computational modeling to understand how brains process social information under naturalistic conditions. Recordings from the macaque mid-STS reveal that the region tracks spatial context, behavioral state, and social prediction errors during natural behavior.

All publications

I build machine learning tools that enable quantitative study of human and animal behavior at scale — from primate face analysis and pose estimation to clinical video understanding and movement science. These tools serve as both scientific instruments and benchmarks for evaluating AI generalization.

All publications

In a new research direction, I study how large vision and language models represent the world, and when our interpretability methods actually recover that — building benchmarks that test interpretability and oversight claims against ground truth instead of taking them on faith.

All publications