How can we design reliable decisions directly from data, without committing to fragile parametric models?

The increasing availability of large-scale data, real-time sensing, and computational power has revived a fundamental question in control: can we replace explicit models with representations of system behavior?

My recent work explores data-driven control through geometry, optimization, and behavioral systems theory. Bypassing parametric model identification, one can design predictive and decentralized controllers directly from data while retaining formal guarantees.

Robustness is a foundational concept in control engineering: a behavior is robust if it persists under the effect of exogenous perturbations or parametric uncertainty.

While robust stabilization of equilibria is well understood, the systematic design of robust non-equilibrium behaviors remains far less developed — despite their ubiquity in physics, biology, robotics, and networked systems.

My work advances a theory of interconnection for systems that operate far from equilibrium, making non-equilibrium behavior amenable to systematic design through control and optimization.

With the rapid advance of computation and artificial intelligence, mathematical models of dynamical systems play an increasingly central role in analysis and design. Yet higher fidelity often comes at the price of high dimensionality, making simulation, control, and interpretation difficult.

Model reduction and system identification address this tension: the former by simplifying while preserving structure, the latter by building models directly from data.

My work investigates these questions through geometric and systems-theoretic methods, seeking scalable and structure-preserving representations that retain stability and robustness.