Research vision
My program combines resolved numerical simulations, complex industrial multiphase flows, experimental and production evidence, and probabilistic learning with Bayesian optimization. Conservation laws, uncertainty, provenance, and validity remain visible when a model informs action.
Geometric VOF and LENT methods preserve interfaces, mass-momentum consistency, force balance, wetting kinematics, and conservation on unstructured meshes.
Microcavity filling, dynamic wetting, capillary bridges, viscoelastic interfaces, surfactants, and interfacial mass transfer connect method development to difficult applications.
High-speed microchannel images, automated contact-angle and interface extraction, tribometer measurements, and process data expose missing physics and real variation.
Gaussian-process surrogates choose informative simulations and experiments while uncertainty, feasibility, and V&V gates remain part of the decision.
Research program
The microfluidics line is supported by peer-reviewed publications, software, and public datasets. Bayesian optimization is an active program supported by public workflows, verified CFD evidence, an industry-funded doctorate, and a scientific conference contribution. Validity-aware monitoring is the next programmatic step.
Four Bosch–TU Darmstadt studies connect stable capillary-flow simulation, validated dynamic-wetting measurements, geometry-resolved pinning, and three-dimensional microcavity filling regimes.
See the four-study program →Gaussian-process surrogates decide where another simulation or experiment has the highest value, then support constrained operating decisions with predictive uncertainty.
See the program →Building on verified CFD, experimental image analysis, and uncertainty-aware surrogates, the next program develops monitors that detect departure from calibrated operating regimes and determine when new simulation or experimental evidence is required.
Scientific leadership
Independent group leadership, first supervision, collaborative funding, and maintained software are treated as one research system.
doctorates as primary supervisor, with five doctoral projects fully industry funded.
SFB 1194 Principal Investigator of the central Z-INF project in the final phase; contributor to successful B01 and B02 proposals.
NFDI4Ing Led the research-software-development measure and advanced reusable data/software practice.
Collaborations Published and ongoing work with Bosch, HPE, Rimac, and the University of Manchester, documented through shared papers, software, data, supervision, and public scientific contributions. View the public record →
Long horizon
High-resolution simulations generate structured evidence and expose missing physics. Experiments correct models and quantify real variation. Probabilistic learning makes uncertainty operational. Agentic workflows accelerate implementation, testing, documentation, and evidence assembly; verification, validation, provenance, and human judgment remain the acceptance layer.