Active Matter

Model-Free Measurement of Local Entropy Production and Extractable Work in Active Matter


Fast Generation of Spectrally-Shaped Disorder

Protein Fitness Landscapes

Predicting and Interpreting Protein Developability via Transfer of Convolutional Sequence Representation

About us

We are a team of problem solvers dedicated to discovering the fundamental principles governing the behaviour of both natural and artificial systems through the development of new mathematical and computational approaches to characterize/engineer order, function, and learning in complex systems.

Our team is composed of theoretical physicists, applied mathematicians, and theoretical chemists doing true interdisciplinary theoretical research. We seek to answer questions like: What is the mechanism by which cortical areas in the brain communicate? Can we engineer disorder in material systems to tune their function? What is the basic mechanisms by which stochastic optimization leads to optimal solutions in machine learning and statistical physics models? How can we extend thermodynamic concepts to systems far from equilibrium? How can we learn a representation of chemical space that rivals the representational capacity of machine learning models for language and vision?

We have introduced computational and theoretical frameworks that revealed striking regularities in physical systems for which one would expect none, and that enabled us to answer questions that could have never been approached with existing techniques. While we relish mathematical and computational sophistication, we are visual thinkers seeking ever simpler representations of complex concepts, and more intutive approaches to solving difficult problems.

Latest news

  • 2/21/2024 Stefano is a co-recipient of the Chan-Zuckerberg Initiative (CZI) Collaborative Pair Pilot Project Award with Prof. André Fenton from NYU's Center for Neural Science.
  • 1/29/2024 Stefano becomes Affiliate Faculty of the NYU Center for Data Science.
  • 12/15/2023 New major update to FReSCo by Aaron and Mathias, now generalized to arbitrary fields OR point patterns, in real and Fourier space, using either periodic OR free boundary conditions, all at high speed O(NlogN). With these variants, we can design a field with desired correlations while obeying other constraints, e.g. conserved total mass, or point patterns with Bragg-peak-like features at arbitrary positions. Thus we obtain 2d and 3d quasicrystals nondeterministically from simple optimization!
  • 11/25/2023 Stefano participates as a guest on the program "Si può fare" by Radio24 (radio station of the Italian financial newspaper of record "Il Sole 24 Ore", 2.2M listeners per day) to talk about Artificial Intelligence and foundation models for materials discovery. The interview is availble from minute 20 of this podcast.
  • 11/18/2023 Stefano delivers a talk as part of the minisymposium on “Suppression and Variability in Visual Cortex” at the Society for Neuroscience annual meeting in Washington D.C., titled "ORGaNICs: A Recurrent Circuit Theory of Normalization"
  • 11/13/2023 Stefano delivers a colloquium at the Center for Computational Neurosciene, Flatiron Institute, titled: "Recurrent Circuit Theory of Cortical Communication"
  • 10/09/2023 Stefano receives the 2023 Interdisciplinary Early Career Scientist Prize from the International Union of Pure and Applied Physics (IUPAP) "for groundbreaking contributions to the understanding of the statistical mechanics of active and amorphous systems via the development of uniquely original approaches for quantifying order, entropy and entropy production in systems far from equilibrium, including granular and active matter, neural networks and biological systems."
  • 09/29/2023 Shivang presents a spotlight talk and Asit a poster at the 2023 Mathematics of Neuroscience conference.
  • 09/19/2023 U.S. Senate Majority Leader Chuck Schumer and U.S. Senator Kirsten Gillibrand announce the FERMat project in a joint press-release.
  • 09/15/2023 The Martiniani Lab receives a $4.5M NSF 5-year GOALI award to develop FERMat, a foundation model for molecular and material property prediction, and ML interatomic potentials for modeling atomic behavior. This project will be performed in collaboration with a team of 6 co-PIs at U. Minnesota, U. Florida, and BYU, and 1 industry co-PI at Amazon Web Services.