spectral-cognitive-labor CI scres-ia CI HeliOS CI

Undergraduate in Applied Mathematics & Computer Science at Universidad del Rosario (Bogotá, Colombia), on a national merit scholarship. I am working toward a research career in NeuroAI and computational neuroscience — at the intersection of reinforcement learning, collective behavior, and computational models of coordination under uncertainty, with current work moving toward the Free Energy Principle / Active Inference.

My work sits between mathematics, machine learning, and the study of how agents — biological or artificial — coordinate and adapt under uncertainty. I try to keep what I build reproducible, honestly scoped, and grounded in theory.

• Collective behavior & multi-agent coordination with Prof. Edgar Andrade-Lotero (UC Davis) — behavioral experiments built on PsyNet, and a spectral / graph-theoretic reanalysis of self-organized division of cognitive labor. A geometric signal computed from the first five shared rounds predicts later axial specialization (AUC_LOOCV = 0.86 when combined with early behavioral metrics; see spectral-cognitive-labor).
• Supply-chain resilience as a learned capability with Prof. Alexander Garrido — PPO-based reinforcement learning over discrete-event simulations of disruption. PPO outperforms the best static policy under severe + stochastic disruption (500k timesteps ×ばつ 5 seeds, frozen benchmark backbone; manuscript in preparation — see scres-ia).
• Incoming participant — Neuromatch Academy 2026, NeuroAI track.

Python · PyTorch · Gymnasium / Stable-Baselines3 · NumPy / SciPy · SimPy · NetworkX · C · LaTeX · reproducible pipelines (Make, pytest, CI)

• Bogotá, Colombia
• Academic: thomas.chisica@urosario.edu.co
• Personal: chisicathomas@gmail.com