Unlocking Autonomy: Event-Driven Linked Data for Scientific Workflows

Abstract: Recent advances in AI-driven science reflect a closing gap between automated instrumentation and computational workflows. Truly transformative progress toward autonomous operation and discovery, however, requires connection across a complex and stateful, event-driven research landscape spanning diverse, distributed laboratories and facilities.

This presentation draws on examples from the AIMD-L programmable cloud laboratory and the NASA IMQCAM STRI autonomous digital twin for additive manufacturing to show how a semantic, event-driven data layer is the thread connecting samples, measurements, workflows, models and decisions across distributed experimental and modeling environments.

These semantic connections make it practical to automatically construct the linked data structures that learning, inference and decision-making require. In this sense, autonomy and transformative AI in science is not just an automation and algorithm problem, but it is a data problem, and one that can now be solved.