Senior AI Research Scientist
Ipazia SpA
Research on language model interpretability, long-horizon planning, and compound AI systems. Work includes Hyperdimensional Probe, SokoBench, and auditable information extraction architectures.
Curriculum / Carlo Nicolini, Ph.D.
About me & curriculum
AI research scientist and physicist working across statistical mechanics, information theory, and machine learning.
I’m a physicist and AI research scientist. My background is in computational neuroscience: during my doctoral and postdoctoral work at IIT, I studied community structure and spectral entropy in brain functional connectivity networks. I now work on language model interpretability and reasoning, alongside scientific software and portfolio optimization.
Based in Milan / Rovereto, Italy. Fellow of the Thinking About Thinking Society and board member at skfolio Labs.
Experience
Co-author & core maintainer · Board member
skfolio & skfolio Labs
Scientific software for portfolio optimization and risk modeling, online convex optimization, compiled solver acceleration, and research strategy.
Senior Quantitative Data Scientist & Data Engineer
Prometeia
Quantitative modeling, distributed analytics, systemic financial risk, and graph-based anti-money laundering pipelines.
Postdoctoral Research Fellow
Istituto Italiano di Tecnologia
Statistical mechanics and spectral entropy approaches to complex networks and brain functional connectivity. Open-source scientific software in Python and C++.
Scientific Programmer & Research Assistant
IIT & University of Trento
Real-time stereoscopic tracking, computational psychophysics, machine learning optimization, and scientific computing.
Education
Ph.D. in Applied Physics
University of Verona
Nanoscience and complex systems. Dissertation on the modular structure of brain functional connectivity networks, conducted at IIT.
M.Sc. in Physics and Biomedical Technologies
University of Trento
110/110 cum laude. Monte Carlo simulations and stochastic optimization for inverse radiation treatment planning.
B.Sc. in Physics
University of Trento
Nonlinear dynamics and feedback systems in cardiovascular pressure regulation.
Research & tools
AI interpretability & reasoning
Vector Symbolic Architectures, neural probing, latent representation geometry, long-horizon planning, probabilistic language programming, and auditable RAG.
Computational neuroscience
Brain functional connectivity, community detection and resolution limits, multiscale network analysis, spectral entropy, and computational psychophysics.
Statistical physics & mathematics
Spectral entropy, graph Laplacians, spanning forests, community detection, information theory, convex optimization, and regret minimization.
Scientific software & systems
Python, C++, PyTorch, JAX, scikit-learn, NumPy, SciPy, CVXPY, distributed computation, and open-source maintenance.
Explore the publications →