Student Research Programs

Sponsors in 2026 include the following:

PROJECT 1:

• Project Title: Identification and Optimization of chloride storage from green power

Safe mass storage of chloride in ionic fluids holds significant potential for temporal energy shifting, enabling optimal use of fluctuating green energy sources. This project focuses on parameter identification and optimal control methods for a chemical reactor system that stores chloride in an ionic fluid. The system integrates fluid flow, species and heat transport, and chemical reactions. Building on a preliminary sensitivity analysis, the goal is to develop model simplifications that facilitate parameter identification from experimental data and to optimize the reactor for both efficiency and throughput. The project requires mathematical techniques such as PDE calculus, finite difference or finite element discretizations, and gradient-based optimization methods. Proficiency in Python and/or MATLAB/Octave is advantageous. By bridging mathematical modeling, numerical simulation, inversion, and optimization, this project offers novel insights into technological aspects of the energy transition.

PROJECT 2:

• Project Title: Task-Specific Foundation Models for Cryo-Electron Tomography

Cryo-electron tomography (cryo-ET) enables visualization of macromolecular complexes in their native cellular context, but interpretation remains challenging due to high noise levels, missing information, and lack of ground-truth data. In this project, we take advantage of a previously established foundation model called CryoSiam for self-supervised representation learning in cryo-ET. CryoSiam is trained to learn a hierarchical representation of tomographic data using a synthetic dataset that systematically models realistic image acquisition parameters. CryoSiam supports model transfer to experimental data without fine-tuning and supports key aspects of cryo-ET data analysis, including tomogram denoising and semantic segmentation of subcellular structures, and macromolecular detection and identification across both prokaryotic and eukaryotic systems. In this project, we aim to modify CryoSiam toward more task-specific outputs like macromolecular orientation estimation, particle picking, and membrane segmentation. For this, we will use our own simulator software called PolNet to generate a diverse dataset including 150 types of macromolecule and other cellular structures. Then, we will integrate that synthetic data with unannotated experimental data to fine-tune the model for the aforementioned tasks. The expected output is a robust model tailored for the Cryo-ET workflow.

PROJECT 3:

• Project Title: Mathematical Models of Deformation and Articulation in Model to Image Registration

This project investigates the mathematical foundations of deformable 2D–3D geometric registration for orthopedic analysis.
Specifically, we aim to align 3D bone geometries, either reconstructed from CT or obtained from a statistical shape prior, with clinically acquired X-ray images, while explicitly modeling biomechanically plausible articulations (joints) and partially rigid (bone) plus non-rigid (soft tissue) shape deformations.
The core challenges include solving an ill-posed inverse problem that couples projective geometry, differential shape analysis, and constrained optimization. The project will enable quantitative assessment of patient-specific kinematics under varying load-bearing conditions. Clinical validation will be performed using data obtained in collaborations with Charité Berlin, Arcus Clinic Pforzheim, and University of Brescia, Italy.