About me

I am a postdoctoral researcher at the Technical University of Berlin in the research group Differential Equations, led by Prof. Etienne Emmrich.

From 2027, I will serve as a principal investigator of DOC-TWIN (MATH+) and scientific co-lead of LenhTVHM (DFG–NSFC), two funded research projects on duality-based error control for digital twins and learning-enhanced iterative methods.

From March to September 2026, I was a Distinguished Fellow at George Mason University (Fairfax, VA, USA), hosted by Prof. Harbir Antil.

Previously, I held a postdoctoral position at the University of Freiburg, where I also completed my PhD under the supervision of Prof. Michael Růžička. From July 2023 to March 2024, I was a Walter Benjamin Fellow of the German Research Foundation (DFG) at the University of Pisa, working with Prof. Luigi C. Berselli on numerical methods for smart fluids.

Open position

We invite applications for a two-year, full-time postdoctoral position at WIAS Berlin within the MATH+ project DOC-TWIN, which I co-lead with Michael Hintermüller and Harbir Antil.

The project focuses on duality-based error control for PDE-constrained optimisation in digital-twin-based structural health monitoring. The position is scheduled to start on 1 January 2027, with remuneration according to TVöD Bund E13.

Please see the full job advertisement (reference 26/22) for details. Applications must be submitted through the WIAS application portal linked in the announcement.

Research profile

My research focuses on the numerical analysis of nonlinear PDEs and nonsmooth variational problems. I combine nonlinear functional analysis, convex duality, and structure-preserving discretisation to develop reliable approximations, computable error estimates, and efficient nonlinear solvers.

Two closely connected areas form the core of my work:

  • Duality-based numerical methods. I use continuous and discrete convex duality to derive primal-dual error identities for a priori and a posteriori analysis, adaptive refinement, and error control for iterative solvers. My work includes explicit flux reconstructions and solvers based on semi-implicit gradient flows and prox-based semismooth Newton methods. Applications include obstacle and Signorini problems, TV minimisation, and digital twins for structural health monitoring. Current directions include subgradient flows, higher-order discretisations, PDE-constrained optimisation, and learning-enhanced iterative solvers.

    Duality connects a priori error analysis in blue, a posteriori error analysis in orange, and iterative solvers in purple. Arrows show the mutual links between the three areas.
    Explore an adaptive Signorini mesh
    Authentic triangular mesh on the square from minus one to one in both directions after 15 adaptive steps. Refinement concentrates along the bottom contact boundary and near the transition at the middle of the right boundary.

    After 15 adaptive steps, primal–dual gap indicators concentrate refinement along the contact boundary and near the boundary-condition transition at (1, 0).

    Signorini problem · Figure 5

    Duality connects analysis and algorithms

    Convex duality provides a common framework for a priori estimates, a posteriori error control, and iterative solution methods.

    Selected papers

  • Analysis and numerics of non-Newtonian and smart fluids. I develop existence theory for nonlinear flow models and establish convergence and error estimates for finite element and discontinuous Galerkin approximations. My work includes variable-exponent and micropolar fluids, evolution equations in time-dependent energy spaces, and generalised pseudomonotonicity methods. Current research addresses fully coupled smart-fluid systems, particularly chemically reacting fluids and models of synovial lubrication.

    Pulsatile flow through a pipe

    Fully developed pulsatile flow: axial velocity in blue, pressure variation in purple. The model allows the material exponent to vary across the pipe.

    Selected papers

Further directions complement this programme:

  • Optimal thermal insulation. I study the optimal distribution of a prescribed amount of insulating material around heat-conducting bodies, accounting for their geometry and convective heat exchange with the surroundings. My work combines variational analysis with finite element approximation and iterative optimisation methods.

    Insulating a re-entry capsule

    Qualitative finite element example on an idealized re-entry capsule. Four material budgets show how the optimal insulation expands from 9% to 71% of the hull.

    Selected papers

  • Neural approximation of parametric PDEs. I study neural approximations of parameter-to-solution maps and their mathematical error analysis. Building on best-approximation results and a priori estimates for the Deep Ritz method, current work combines variational energies with physics-informed residuals for parametric p-Stokes systems, including parameter-dependent material laws, forcing, and geometries.

    Families of PDE solutions

    Forcing, geometry, and the constitutive exponent each generate a different solution family. These surfaces show exact scalar model examples; the orange curves select μ = 0.5.

    Selected paper

  • Quantum-hydrodynamic generative modelling. I develop PDE-based generative methods using Schrödinger dynamics and their Madelung formulation. PDE-constrained optimisation determines an initial phase so that the resulting deterministic dynamics transport a reference distribution towards a prescribed target. Current research addresses scalable phase identification, structure-preserving discretisations, and error control.

    From one peak to two: density evolution

    Four snapshots show how a computed initial phase transforms a Gaussian density towards a bimodal target (T = 0.3). The target is shown separately; all panels share the same spatial, height and colour scales.

    Selected paper