AIchemy’s Monthly Webinar Series – September 2026

We are delighted to welcome you to our AIchemy Hub’s monthly webinar series.

This month’s talks:

Dr. Linjiang Chen – University of Birmingham

Talk Title: Autonomous chemical research with robots and agents

I will introduce our robotic AI chemists, integrating knowledge acquisition, theoretical modelling, machine learning, and automated experimentation [1,2], with catalyst discovery for plastic recycling illustrating the connection between data-driven optimisation and hypothesis-guided exploration [3].

I will then describe our robotic AI chemistry research infrastructure spanning more than 2,600 m² with 605 registered workstations. Modular architecture and reusable skills connect heterogeneous instruments, computational tools and decision models while preserving workflow traceability [4].

However, infrastructure alone does not ensure autonomy. Across 4,608 agent trials, only 3.3% produced expert-assessed executable workflows; the best configuration achieved 28.1%. A five-round adaptation test revealed local parameter adjustments but no workflow-level replanning or analytical-method redesign [5].

To address the operational side of this gap, we developed a computable representation of the physical laboratory, combining typed research objects, capability-bound operations, and a compositional workflow algebra [6]. Plans become programs over evolving sample and container states, checked through structural analysis and stateful simulation before dispatch. Four workflow examples illustrate this verification, which establishes consistency with encoded constraints—not chemical success. Together, these studies highlight why autonomous chemistry requires a computable, testable interface between scientific reasoning and physical action.

References

1. Zhu, Q. et al. Natl. Sci. Rev. 9, nwac190 (2022). DOI: 10.1093/nsr/nwac190.
2. Song, T. et al. J. Am. Chem. Soc. 147, 12534–12545 (2025). DOI: 10.1021/jacs.4c17738.
3. Yu, Y. et al. J. Am. Chem. Soc. 148, 4635–4644 (2026). DOI: 10.1021/jacs.5c20630.
4. Li, X., Chen, L. et al. ACS Nano 20, 24593–24603 (2026). DOI: 10.1021/acsnano.6c15286.
5. Guo, L. et al. Stress-testing large language model agents in a robotic chemistry laboratory. arXiv:2607.23045 (2026). Preprint.
6. Li, X. et al. A computable representation of the physical laboratory enables verifiable workflows. arXiv:2609.03621 (2026). Preprint.

Lyubomir Kootopanov – University of Liverpool

Talk Title: From In Silico Design to Automated Synthesis: An AI-Driven Framework for Late-Stage Functionalisation

Self-driving laboratories guided by machine learning algorithms hold the potential to significantly accelerate discoveries in the chemical sciences. However, to fully exploit the opportunities offered by autonomous laboratories, medicinal chemistry faces a major bottleneck – not only in the physical preparation of molecules, but also in computer-based retrosynthetic planning. Current algorithms favour well-established chemistries, whereas newer, potentially more efficient reactions are under-prioritised because they are less explored in the literature. We present a flexible data-driven framework to guide high-throughput late-stage functionalisation of drug compounds. This pipeline generates diverse therapeutic candidates in silico with promising properties and synthesisability via user-defined late-stage functionalisation reactions with predicted conditions. Our results suggest that data-driven methods can be utilised to establish a robust and standardised workflow for the autonomous generation of promising drug-like molecules that can be efficiently accessed via innovative reaction pathways.