Events

CDS Colloquium: Building Reliable Knowledge Across Disciplines: From Supernovae to the Systems of Science

Speaker: Wolfgang Kerzendorf

Location: 60 Fifth Avenue, Room Room 150

Date: Friday, November 21, 2025

Progress on today’s most demanding scientific questions, whether understanding stellar explosions or confronting climate and biomedical complexity, depends on extracting reliable knowledge from data through simulation-based inference. Such work is inherently interdisciplinary, drawing on physics, statistics, and machine learning. My research builds this bridge at two complementary levels: developing trustworthy inference frameworks for complex phenomena and creating the meta-scientific infrastructure that sustains interdisciplinary collaboration. In the first part, I present a new framework for astrophysical inference, where we reconstruct supernova explosions from spectral time series by coupling probabilistic neural-network emulators with Bayesian sampling. This approach enables millions of model evaluations that would be impossible with direct Monte Carlo radiative-transfer simulations, yielding new insights into the composition and energetics of stellar ejecta. In the second part, I turn to the practice of science itself. We develop open-source, machine-learning-driven infrastructure to make research more transparent and scalable: a neural global researcher registry that uniquely identifies contributors across disciplines, a data-driven peer-review platform that matches proposals to expertise efficiently and fairly, and metadata-extraction tools that reveal how research facilities and collaborations are actually used. Together, these efforts demonstrate how interpretable machine learning can advance both our understanding of the universe and the collaborative systems that make such understanding possible.

Bio: Dr. Wolfgang E. Kerzendorf is an Assistant Professor in the Departments of Physics & Astronomy and Computational Mathematics, Science & Engineering at Michigan State University, where he joined the faculty in 2019. His research connects the physics of plasmas and radiative transfer in stellar explosions and supernova ejecta with data-driven and machine-learning methods. As lead author and principal investigator of the open-source radiative-transfer code TARDIS, he develops community tools that couple atomic physics, radiation transport, and inference techniques to interpret the spectra of astrophysical transients. Dr. Kerzendorf’s group uses emulation, probabilistic inference, and modern statistical frameworks to map the relationships between theory and observation, revealing how progenitor structure and explosion physics shape observed spectra. His work spans detailed microphysics and large-scale inference, creating bridges between plasma modeling, radiative transport, and data science. Beyond astrophysics, he leads the DeepThought Initiative, a computational meta-science effort aimed at enhancing the process of scientific discovery itself. This work develops algorithmic and data-centric approaches to improve how science is conducted, reviewed, and connected in an era of exponential growth in both publications and researchers. It also focuses on strengthening the robustness and transparency of the scientific process at a time when public trust in science faces increasing challenges. Dr. Kerzendorf earned his Vordiplom in Physics from Universität Heidelberg and his Ph.D. in Astronomy & Astrophysics from the Australian National University in 2011. He held postdoctoral appointments at the University of Toronto, the European Southern Observatory, briefly as a senior researcher at NYU before joining MSU.