Abstract

Contributed Talk - Splinter EScience   (MW-2235)

AI Agents for Reproducible Astronomy: DRP-Hub as a Science Accelerator

Arman Khalatyan
Leibniz-Institut für Astrophysik Potsdam (AIP)

Astronomy is shifting toward conversational interfaces for scientific data, where AI agents interpret natural-language goals and handle data access, workflow orchestration, and analysis execution. This talk shows how such agents, integrated with FAIR-compliant infrastructures, improve accessibility, transparency, and reproducibility in data-intensive science. The centerpiece is DRP-Hub https://drphub-p4n.aip.de, a platform unifying datasets, workflows, and AI agents, now integrated with REANA to generate and execute portable, containerized workflows across distributed infrastructures. Developed with PUNCH4NFDI and PhysicsLLM, this work brings FAIR research infrastructure together with domain-specific AI. Using Gaia survey data as an example, we demonstrate conversational interaction with large datasets while preserving scientific quality and provenance — pointing toward a future where every astronomer has a personal AI science assistant.