University of Wisconsin–Madison Data Science Institute (DSI) Research Scientist Matthew Feickert is one of two inaugural recipients of the US Research Software Sustainability Institute (URSSI) Early-Career Fellowship. This program supports emerging researchers whose work advances scientific software development practices.
As an URSSI Fellow, Feickert will focus on using new technologies and standards from the broader scientific software ecosystem to identify best practices and design standards for reproducible machine learning workflows for scientists. These practices will be turned into an open source course contributed to The Carpentries, and this new course will be taught at a national-level workshop held at UW–Madison and as a tutorial at the 2025 international SciPy conference.
The URSSI Fellowship funds research projects focused on improving current disciplinary or domain practices in scientific software development, with a focus on the growing use of AI, emerging best practices in sustaining scientific software, and education research to improve design and delivery of software training.
Feickert is a particle physicist working on the ATLAS experiment at the Large Hadron Collider, and he leads the Analysis Systems focus area of the Institute for Research and Innovation in Software for High Energy Physics (IRIS-HEP). Using and improving research software has been a strong focus throughout his research career, and he knows firsthand the challenges of analysis reproducibility.
“Modern scientific analyses are complex software and logistical workflows that may span multiple software environments and require heterogenous software and computing infrastructure,” says Feickert. “As researchers, we want to do our best to make our work reproducible and reusable, both for our current and future colleagues, but this has historically required software engineering experience that not all scientists have the background for. Adding requirements for machine learning workflows was a whole other level of complexity, but with recent advancements, the complexity barrier is lowered.”
Along with developing new tools and technologies, Feickert has explored partnerships between academia and industry. He says there is an opportunity to make fully reproducible software environments for machine learning and hardware-accelerated scientific workflows achievable with high-level semantics targeted at researchers.
Feickert hopes that his work will impact research beyond UW–Madison. “I’m excited to not only bring these improvements to my own physics research and to share them with my colleagues at the DSI, but to share these technologies and practices with researchers around the globe as ‘sensible defaults’ that lower barriers to use and accelerate research.”