"This tutorial introduces self-driving laboratories (SDLs) and software tools required to operate them, focusing on orchestration, optimization, and closed-loop experimentation. Through a demo-driven, hands-on format, attendees will work through notebooks or hosted tutorials to explore key concepts in practice. We begin with IvoryOS, which demonstrates how abstraction and a drag-and-drop interface produce reproducible workflows and reduce barriers for experimental teams. We then move to optimization. NIMO will then present a unified interface for invoking a variety of algorithms for closed-loop experimentation. NextSample turns Bayesian optimization into a customizable app layer for experiment-specific planning workflows. Finally, we introduce “legacy displacement,” a conceptual software development framework for upgrading SDL software without interrupting ongoing operation. The tutorial concludes with a panel discussion to interrogate limitations, clarify implementation, and identify opportunities for adoption"

Integrating laboratory equipment with heterogeneous automation readiness levels remains a significant practical barrier to building Self-Driving Labs (SDLs) and Materials Acceleration Platforms (MAPs). Many laboratories rely on legacy instruments, vendor-locked software, or poorly documented interfaces, yet wish to incorporate these devices into reproducible, closed-loop experimental workflows. This tutorial provides a structured overview of four “automation readiness levels” (ARLs) of laboratory equipment, ranging from devices with no software interface to instruments offering fully supported APIs or SDKs and presents strategies for integrating each tier into automated workflows. The objectives of the tutorial are to (1) equip participants with a conceptual framework for assessing instrument readiness, (2) demonstrate practical integration methods through live coding examples, (3) highlight common pitfalls and robust design principles, and (4) share real-world case studies drawn from building operational SDLs. Attendees will gain hands-on experience with tools such as GUI automation packages, serial-command protocols, and vendor APIs, as well as broader methodological insights for planning, implementing, and troubleshooting automation pipelines.

This workshop will focus on low-cost alternatives to commercial automated experimentation systems and the philosophy underpinning these tools. Designed to be affordable (SG$ 1000–2000), straightforward to programme, and highly customisable, these systems can be fitted into a standard fume cupboard and used safely with reagents common in synthesis and characterisation. Whilst not intended to replace high-fidelity platforms, low-cost tools can augment the high-fidelity systems, improve throughput, simplify many routine laboratory tasks and serve as valuable teaching and demonstration instruments for young researchers.

Artificial intelligence is evolving from static predictive models into agentic systems built on large language models that can reason, plan, and coordinate multi-step scientific tasks. These systems act as embedded digital collaborators, capable of orchestrating simulations, data analysis, experimental design, and knowledge integration. The first part of this two-part workshop establishes the theoretical and architectural principles behind agentic AI, including a broad educational overview of the underlying concepts, design patterns, how scientific problems can be decomposed, how search complexity can be managed, and how autonomous agents can be designed to support scalable and interpretable discovery across chemistry, physics, and biomedical domains.

Building on the theoretical foundations from Part 1, the second section focuses on practical implementation of agentic AI systems in scientific workflows and laboratory environments. Through interactive walkthroughs and hands-on exercises using lightweight open-source tooling, participants will learn how to construct simple agentic pipelines, configure tool-calling workflows, and build minimal multi-agent systems for scientific tasks. The workshop will also extend these concepts toward real experimental infrastructures and scientific instrumentation. Using case studies centered on NMR spectroscopy, we demonstrate how computer-use agents can interact directly with GUI-based laboratory software, accumulate reusable procedural knowledge, and bridge legacy instruments with emerging self-driving laboratory paradigms.

The use of AIs and self-driving labs (SDLs) in materials discovery can help us develop better materials that help society, but it also raises ethical dilemmas, including possible weaponization. Self-driving laboratories (SDLs) hold transformative potential for accelerating scientific discovery yet realizing that potential safely demands rigorous and multi-disciplinary approaches to reduce harm. If we want to think and build sustainably, we need to think about research design and outcomes intergenerationally, and not just short term. One way to do this is through harm reduction. The session will also examine the institutional and ethical dimensions of harm reduction, from personal responsibility to the governance of increasingly autonomous research infrastructure. Participants will leave with practical tools and a shared vocabulary for integrating harm reduction into SDL design, operation, and oversight contributing to a safer and more trustworthy foundation for the next generation of automated science.