Research
From the quantum dynamics of single molecules to AI systems used by many people – the common thread is making complex simulations and creative tasks fast and accessible.
Machine learning for molecules and materials
Designing new materials and molecules – for medicines, electronics or solar energy – requires understanding how they behave at the atomic level: how molecules pack into crystals, or how they respond to heat or light. Simulating these properties with quantum mechanics is extremely accurate but painfully slow.
Machine-learning models trained on quantum-chemistry data learn these patterns and reproduce quantum-level accuracy at a tiny fraction of the cost. This makes it possible to screen large numbers of compounds and light-driven processes, bringing powerful simulations closer to real-time decision-making in chemistry and materials science.
Machine learning for graphic design
Creating effective visual content – for presentations, social media or branding – needs both design skill and time. Machine learning can assist and automate parts of the creative process, generating images, layouts and text from simple descriptions.
Large language models and image-generation models such as diffusion models power features that suggest layouts, rephrase text, edit images or create new visuals from scratch, helping individuals and teams produce professional content in seconds, even without design experience.
Excited states and photodynamics
What happens in the femtoseconds after a molecule absorbs light decides whether DNA is damaged, whether a solar cell works, or whether a photoswitch flips. I developed simulation methods for these non-adiabatic processes and co-developed the SHARC surface-hopping package.
Combining these simulations with neural networks made it possible to run excited-state molecular dynamics for much longer times and larger systems than before – an early step towards today's machine-learned interatomic potentials.
Laser–molecule interactions
Extremely short, intense laser pulses act like a high-speed camera made of light: lasting only a few femtoseconds (10⁻¹⁵ s), they are fast enough to follow how atoms and electrons move during a reaction.
Quantum-dynamical simulations show how a shaped laser field interacts with a molecule – how it can break, form or rearrange bonds – and how light can be used to control chemical reactions.