Stefan Heinen up in Switzerland, playing soccer for over 20 years up to the second league. During college, he discovered the trombone and became an active member of a “Guggenmusik” group called Caracas (caracas.ch). He then went to Zürich where he completed his Bachelor’s and Master’s degrees in Chemistry at ETHZ. There, he also helped found the Societas Studentium Vallesanorum Turici (SSVT, http://ssvt.ch/), a student organization that supports students from Wallis (his home canton) in their undergraduate journey, connecting them with the right people and keeping them in touch with home through events like Raclette parties. Subsequently, he pursued a PhD in machine learning applied to quantum chemistry at the University of Basel. Following his PhD, Stefan worked as a Postdoc at the University of Vienna and later joined the Vector Institute as a Postdoctoral Fellow. His research primarily focuses on developing machine learning algorithms, particularly Kernel Ridge Regression, for studying chemical reactions in quantum chemistry. He also investigates computational costs in this field, specifically optimizing scheduling on high-performance computers (HPC) using machine learning. Currently, he is am finalizing a paper on multi-level learning, where he utilizes lower levels of theory (e.g., Hartree Fock) to minimize reliance on more expensive post Hartree Fock methods like CCSD(T) while maintaining accuracy. Additionally, he is working on their “in-house” machine learning code written in Fortran with a Python overlay. The improvements involve incorporating additional representations, refining the wrapper, and overall code cleanup.
Research Interests
- Quantum Machine Learning
- Machine Learning in Chemical Compound Space
- Machine Learning
- Quantum Chemistry
- Coding
Highlights
- Machine Learning Computational Cost of Quantum Chemistry (10.1088/2632-2153/ab6ac4)
- Toward the design of chemical reactions: Machine learning barriers of competing mechanisms in reactant space (10.1063/5.0059742)
- Quantum machine learning at record speed: Many-body distribution functionals as compact representations (10.48550/arXiv.2303.16312)