Supercomputing und E-Science

Postdoctoral Researcher on LLM-based Agentic Systems for Scientific Workflows (w/m/d)

The Leibniz Institute for Astrophysics Potsdam (AIP) is dedicated to astrophysical questions ranging from the study of our sun to the development of the cosmos. Research focuses on cosmic magnetic fields and extragalactic astrophysics as well as the development of research technologies in the fields of spectroscopy, robotic telescopes and e-science. The AIP carries out its research mission within the framework of numerous national, European and international collaborations. The institute is the successor to the Berlin Observatory, founded in 1700, and the Astrophysical Observatory Potsdam, founded in 1874, which was the first institute in the world to dedicate itself explicitly to astrophysics. The AIP has been a member of the Leibniz Association since 1992. Around 200 employees work at our location in the middle of a beautiful park landscape in Potsdam, not far from Berlin.

The Supercomputing and E-Science section at AIP invites applications for a pivotal role within its team:

Postdoctoral Researcher Position: LLM-based Agentic Systems for Scientific Workflows (w/m/d)

We are seeking an exceptional researcher with proven expertise in Large Language Models (LLMs), advanced AI systems development, and scientific data management to drive the Physics-LLM project, which is strategically funded by ErUM-Data/BMFTR: Research Data Management.

About the Project

The Physics-LLM project is dedicated to advancing next-generation, LLM-driven agentic systems that will fundamentally transform how scientists interact with complex data, automate intricate workflows, and conduct cutting-edge computational research. Operating within the ErUM-Data framework, the project’s mission is to democratize access to sophisticated scientific workflows through intuitive natural language interfaces, thereby enhancing the reproducibility, efficiency, and accessibility of research processes.

As a Postdoctoral Researcher, you will work under the direct supervision of Dr. Arman Khalatyan at the AIP, PI of LLM-based Agentic Systems. This critical work package is structured around three main pillars: core agentic system development, seamless domain knowledge integration, and robust multi-community deployment. Your responsibilities will encompass the design and implementation of core components of the agentic LLM system, with a strong focus on advanced reasoning, memory integration, and autonomous task execution capabilities. You will work on developing, researching and orchestrating sophisticated multi-agent systems to manage complex scientific workflows, integrating specialized domain-specific knowledge bases. The project is a close collaboration between German universities and HPC centers.

Required Qualifications Applicants must hold a PhD (or possess equivalent research experience) in Astrophysics, Physics, Computer Science, Machine Learning, or a closely related field. A strong background or demonstrated interest in artificial intelligence, natural language processing, or data-driven scientific research is essential. Candidates should possess good programming skills, specifically in Python, and be familiar with modern machine learning tools and frameworks.

A solid understanding of scientific data management principles and workflow concepts, coupled with exposure to distributed or cloud-based computing environments, will be highly advantageous. Desired Qualifications Experience with open-source large language model ecosystems such as Hugging Face, Ollama, or vLLM is highly desirable. Additional familiarity with reinforcement learning, autonomous or multi-agent systems, and workflow orchestration platforms such as REANA or Apache Airflow would be a significant asset.

What We Offer

Initial Appointment: 3 years (with potential for extension contingent upon performance and project continuation),
Start Date: Negotiable, ideally early 01.2026
Salary: Competitive remuneration package at the German academic standard level (TV-L E13, depending on qualifications and experience)
Benefits: Comprehensive social benefits package including company pension (VBL), and flexible working arrangements
Work Environment: A highly collaborative, international team environment with access to state-of-the-art research infrastructure
Location: Potsdam, Germany

We particularly value diversity and we encourage applications from women. Preference will be given to applicants with disabilities if they possess equivalent professional qualifications and abilities. For detailed inquiries regarding the position, please contact the hiring team via the following email address:

bewerbung-2025-11@aip.de

Veröffentlicht
06. November 2025
Download
PDF
Abteilung
Kontakt
Online-Bewerbung
E-Mail