Postdoctoral position in simulation-based inference for particle physics

Uppsala University, Disciplinary Domain of Science and Technology, Faculty of Mathematics and Computer Science, Department of Information Technology

Are you interested in working with machine learning and simulation-based inference for searches for dark matter (or other “invisible” new physics signals) at the Large Hadron Collider, with the support of competent and friendly colleagues in an international environment? Are you looking for an employer that invests in sustainable employeeship and offers safe, favourable working conditions? We welcome you to apply for a postdoctoral position at Uppsala University.

The Department of Information Technology holds a leading position in both research and education at all levels. We are currently Uppsala University’s third largest department, have around 350 employees, including 120 teachers and 120 PhD students. Approximately 5,000 undergraduate students take one or more courses at the department each year. You can find more information

about us
on the Department of Information Technology website.

The position is hosted by the Division of Scientific Computing (TDB) within the Department of Information Technology. As one of the world’s largest focused research environments in Computational Science, the research and education has a unique breadth, with large activities in areas such as numerical analysis, mathematical modelling, development and analysis of algorithms, machine learning, optimization, scientific software development and high-performance computing. The division is also an important part of the eSSENCE strategic collaboration on e-science and of the Science for Life Laboratory (SciLifeLab) network, a national research infrastructure with a mandate to enable cutting-edge life science research in Sweden.

The successful candidate will join the Scientific Machine Learning group at TDB and SciLifeLab (Associate Professor Prashant Singh), which develops methods and software for simulation-based inference, generative models and robust machine learning, together with the Theoretical Particle Physics group at the Department of Physics and Astronomy (Professor Stefano Moretti), which works on beyond-the-Standard-Model and dark-matter phenomenology and is a member of the CMS experiment at CERN. Together, the groups offer an international environment with a wide network of collaborators, generous support for conference travel, and access to national HPC resources (NAISS) and local GPU infrastructure.

This Postdoc position is part of the eSSENCE graduate school in data-intensive science. The school addresses the challenge of data-intensive science both from the foundational methodological perspective and from the perspective of data-driven science applications. It is an arena where experts in computational science, data science and data engineering (systems and methodology) work closely together with researchers in (data-driven) sciences, industry and society to accelerate data-intensive scientific discovery. eSSENCE is a strategic collaborative research programme in e-science between three Swedish universities with a strong tradition of excellent e-science research: Uppsala University, Lund University and Umeå University.

In this project we envision a novel cross-faculty collaboration between the Division of Scientific Computing and the Department of Physics and Astronomy, where new methodology for simulation-based inference is developed and applied directly in realistic collider analyses. The postdoc will be hosted at TDB, co-supervised by both groups, and will work at the interface of scientific computing, machine learning and particle physics.

Project description

Searches for dark matter (or other “invisible” new physics signals) at the Large Hadron Collider (LHC) compare high-dimensional collision data against detailed simulations whose likelihood cannot be evaluated, only sampled. Simulation-based inference (SBI) addresses this by training neural networks, such as flow-matching generative models, on simulated events. The project aims to develop efficient, robust and calibrated SBI methods for such searches that account for event selection and systematic uncertainties, and to demonstrate them on realistic searches at scale on national HPC resources. The project will primarily use Monte Carlo simulated data, but there is also the possibility of working with real open data from the ATLAS and/or CMS experiments. The methods are general and applicable well beyond particle physics.

Duties

Research within the project described above, including method development, implementation, large-scale computational experiments on national HPC resources, and publication at machine learning and physics venues. The duties also include presenting results at international conferences, contributing to the group’s open-source software, actively participating in the activities of the eSSENCE graduate school, and tak