Postdoctoral Position - DEMOS and BAT.jl Projects

Postdoctoral Position - DEMOS and BAT.jl Projects

Max Planck Institute for Physics

Garching, Germany

The Max Planck Institute for Physics (MPP) is engaged in fundamental research in particle and astroparticle physics from both experimental and theoretical perspectives. We invite applications for a Postdoctoral position (f/m/d) to work on the development of scientific software, advanced statistical methods, and physics analysis within the framework of the DEMOS and BAT.jl projects. The position also offers opportunities to contribute to research in Bayesian inference and its application to physics in general.

The DEMOS project aims to develop state-of-the-art technology that will make it possible to publish, combine, and preserve statistical models for physics experiments and beyond, and to make them usable across programming languages (C++, Python, and Julia). DEMOS is a new, federally funded consortium of German research institutions and will work in collaboration with international partners to make portable statistical models a reality.

Within the larger DEMOS consortium, the position at MPP will focus on:

• Developing a full implementation of the HEP Statistics Serialization Standard (HS3) in Julia;
• Contributing to the evolution of the HS3 standard to cover a wider variety of physics use cases;
• Developing methods to make machine-learning-based models portable and interoperable;
• Leading the definition of containerized and networked “Models as a Service” (MaaS) and providing the reference implementation in Julia;
• Implementing and publishing several physics analysis use cases using the new framework.

The successful candidate is expected to play a major role in fulfilling MPP's commitments within the DEMOS project. In synergy with DEMOS activities, the candidate should also contribute to the Bayesian Analysis Toolkit in Julia (BAT.jl) and the Julia statistics ecosystem in general. One current area of interest of the MPP BAT group is the adoption and further development of modern Bayesian methods that combine statistics and machine learning.

MPP has long-standing involvement in Belle II physics and neutrinoless double-beta decays (LEGEND experiment), as part of large international collaborations. The successful candidate will collaborate with these groups to apply the developed technology to physics use cases, aiming for publishable results. The candidate may also suggest applications based on their own scientific interests.

A formal requirement for the position is a PhD in physics, computer science, statistics, or a related field.

The candidate should:

• Be interested in statistical inference and methods, with a focus on the Bayesian approach;
• Possess solid software development skills, ideally in more than one programming language. Knowledge of Julia is a plus; willingness to dive deeply into Julia is essential. Basic knowledge of C++ will be beneficial as well;
• If not a physicist by primary training, still have an understanding of physics and the relationship of experiments to mathematical/statistical models.

The primary working language of the projects and the MPP Bayesian analysis group is English.