Abstract

Contributed Talk - Splinter EScience   (MW-2235)

S3-Enabled Reproducible Workflows: Bridging REANA, DRP-Hub, and AI-Driven Data Analysis

Dr. Arman Khalatyan, Enrico Tom Leonhard Strauß
Leibniz-Institut für Astrophysik Potsdam (AIP)

Object storage based on the S3 storage model is becoming a central building block for modern scientific workflows. In this presentation, we highlight recent developments in REANA and DRP-Hub toward native S3 integration, enabling scalable, portable, and reproducible data analysis across distributed infrastructures. We discuss how S3-based storage simplifies workflow orchestration, data provenance, and artifact sharing, making reproducibility more robust and practical. Beyond workflow execution, we explore why S3 is increasingly critical for AI-driven analysis pipelines in astronomy and other data-intensive sciences, where agents, large models, and automated workflows require efficient access to large datasets, intermediate products, and shared knowledge spaces. This convergence of reproducibility, object storage, and AI opens new possibilities for collaborative and autonomous scientific discovery.