Speaker: Marnik Bercx
Affiliation: Paul Scherrer Institut
Abstract
Reproducibility is a core tenet of science, yet many published results cannot be reproduced—eroding trust and wasting valuable time. Even within research groups, new students routinely spend days or months trying to re-create prior results. This talk introduces AiiDA, a scientific workflow manager rooted in condensed-matter theory and, more recently, used to drive experimental work. I will show how AiiDA’s history (provenance) tracking, automation, and robust workflows aim to address common failure points in reproducibility, with concrete examples from real projects. I will also demonstrate how AiiDAlab—a Jupyter-based framework—turns these workflows into accessible GUI applications, lowering the barrier for non-experts. Finally, I will discuss software development in science more broadly: the issues it has, how these tie into the reproducibility problem, and some personal thoughts on what can be done about it.
About Physics colloquium
The Physics Colloquium series hosts a range of speakers from Australia and abroad. The series explores a variety of topics and everyone is welcome to come along. The seminars are open so there is no need to register your attendance.
If you would like to sign up for colloquium announcement emails, you can join the mailing list by sending a blank email to:
- For UQ email addresses: physics-colloquium-others-join@lists.science.uq.edu.au
- For non-UQ email addresses: physics-announce-external-join@lists.science.uq.edu.au
(Note: if you receive physics-all emails, you should already receive these and don't need to sign up again).
