Virtual Laboratories: Domain-agnostic workflows for research

Abstract

Many scientific disciplines have traditionally advanced by iterating over hypotheses using labor-intensive trial-and-error, which is a slow and expensive process. Recent advances in computing, digitalization, and machine learning have introduced tools that promise to make scientific research faster by assisting in this iterative process. However, these advances are scattered across disciplines and only loosely connected, with specific computational methods being primarily developed for narrow domain-specific applications. Virtual Laboratories are being proposed as a unified formulation to help researchers navigate this increasingly digital landscape using common AI technologies. While conceptually promising, VLs are not yet widely adopted in practice, and concrete implementations remain limited.This paper explains how the Virtual Laboratory concept can be implemented in practice by introducing the modular software library VAILabs, designed to support scientific discovery. VAILabs provides a flexible workbench and toolbox for a broad range of scientific domains. We outline the design principles and demonstrate a proof-of-concept by mapping three concrete research tasks from differing fields as virtual laboratory workflows.