From microscopic randomness to macroscopic laws — in particles,
spins, markets, and AI agents.
I study how many interacting random or imperfect components produce
reliable collective behaviour. My tools come from probability,
statistical mechanics, optimal transport, PDE, and data-driven
computation.
I joined UCLA's Department of Mathematics in January 2016. Before
that I held postdoctoral positions at the Max Planck Institute for
Mathematics in the Sciences in Leipzig, Stanford University, and
the Simons Laufer Mathematical Sciences Institute (SLMath, formerly
MSRI). I did my PhD with Felix Otto.
The common theme in my work is interaction. Many simple
objects — particles, spins, random height functions, market orders —
can produce surprisingly stable macroscopic behaviour, and I try to
understand when this happens, how fast it happens, and how robust
the resulting laws are: a noisy microscopic system, on the right
scale, behaves like the solution of a heat equation; random height
functions concentrate on predictable limit shapes; the trades of
many small agents aggregate into a market price.
Lately the same theme has pulled me into AI for mathematics. With
my student Maciej Głuchowski I developed
AGRADE, an automated, multi-agent AI grading tool currently piloted in
the UCLA Mathematics Department. A network of language models,
structured to critique and repair one another, can be far more
reliable than any one of them on its own — which is, in a different
language, the same question that drives much of my mathematical
work: how does reliable macroscopic behaviour emerge from many
unreliable interacting parts?
Interacting particle systems & statistical mechanics
Hydrodynamic limits, convergence to equilibrium, functional
inequalities, phase transitions, metastability, and variational
principles for random maps.
Market microstructure, aggregation of orders into prices,
market-dynamical entropy, non-conservative optimal transport,
operator learning for price impact and execution, and detection
of regime changes in high-dimensional factor models.
Quantitative laws for interacting stochastic systems.
Hydrodynamic limits, functional inequalities, ergodicity, and
macroscopic PDE behaviour emerging from microscopic randomness.
From interacting agents to market prices.
Aggregation of orders, market-dynamical entropy, price impact,
non-conservative transport, and operator learning for execution.
Reliable mathematics from interacting AI agents.
AGRADE, Proof Crystallization, adversarial critique, repair loops,
and multi-agent verification for mathematical reasoning. Also as a
practical introduction: How to get started with AI in teaching
and research (inaugural talk, UCLA AI in Research and
Education seminar, May 2026).
Recent talk:Non-conservative optimal
transport, États de la Recherche SMF: Inégalités
fonctionnelles, CIRM, Marseille, November 20, 2026.
Slides (PDF).
Organizing: AMS Special Session on Lattice Models
of Statistical Mechanics (with Tom Hutchcroft and Yujin Kim), Fall
Western Sectional Meeting, Arizona State University, Tempe,
November 7–8, 2026.
Programme
(PDF).
Short bio (for organizers)
Georg Menz is an Associate Professor of Mathematics at UCLA. His
research studies interacting stochastic systems, using tools from
probability, statistical mechanics, optimal transport, PDE, and
computation. His recent work connects classical problems such as
hydrodynamic limits and functional inequalities with market
microstructure, non-conservative optimal transport,
high-dimensional change-point detection, and AI-assisted
mathematical verification.