NSF PCL-test bed: towards a general platform for AI-driven high-throughput reaction discovery and automation

The National Science Foundation awarded Scripps Research Institute a $4.873 million Cooperative Agreement on August 1, 2026, to establish and operate an AI-guided programmable cloud laboratory node for chemistry within a national network of such facilities. The award is made under NSF Technology, Innovation, and Partnerships (CFDA 47.084).

The node integrates three components into a single platform: advanced robotic and automation capabilities, an AI model broadly trained on chemical reactivity, and an open data format. The platform will conduct high-throughput reaction discovery and optimization using multi-catalytic systems, such as metallophotoredox chemistry, as stress tests for autonomous experimentation. The node will be accessible to vetted users across the country for pharmaceutical, agrochemical, semiconductor, and petrochemical applications.

Beyond platform operation, the award funds workforce development in AI-assisted automated chemistry, targeting technicians, chemists, biologists, materials scientists, and chemical engineers.

Work is performed in La Jolla, California, with a period of performance from August 1, 2026, through July 31, 2030.

Abstract

NSF PCL-test bed: towards a general platform for AI-driven high-throughput reaction discovery and automation artificial intelligence and laboratory automation are poised to transform the discovery and production of medicines, agrochemicals, organic semiconductors, advanced composite materials, and more. However, these capabilities are currently too expensive for broad adoption, and existing platforms are often?optimized?for niche needs. The chemistry PCL node will?establish?an AI-guided laboratory within a national network of programmable cloud labs. The node will integrate three components into a single platform accessible to vetted users across the country: (i) advanced robotic and automation capabilities; (II) an AI model broadly trained on chemical reactivity; and (III) an open data format. The advances in AI-assisted automated chemistry will?benefit?the pharmaceutical, agrochemical, semiconductor, and petrochemical industries, among others. Additionally, the node will train an AI- and automation-empowered workforce that includes technicians, chemists, and chemistry-adjacent scientists such as biologists, materials scientists, and chemical engineers. Together, these scientific and educational contributions will strengthen united states competitiveness and leadership in chemical innovation. This project will build, deploy, and?operate?a programmable cloud laboratory node for chemistry, integrating high-throughput automation, multidimensional analytical instrumentation, and interpretable AI into a closed-loop platform for reaction discovery and optimization. The node will use multi-catalytic systems (e.g.,?metallophotoredox?chemistry) as stress tests for autonomous experimentation. These systems are sensitive to small perturbations, expose the limits of conventional one-factor-at-a-time optimization, and are widely?encountered?across modern catalysis, making them ideal benchmarks for robustness. Alongside reaction discovery, the node will develop a vendor-neutral, web-based platform that unifies robotic execution, data capture, and AI-guided experimental design under a single control plane. IT will host the AI optimization engine, enforce fair-compliant (findable, accessible, interoperable, reusable) metadata standards, and provide secure, role-based remote access for verified academic and industrial users. The node will produce versioned datasets, transparent model registries, and traceable optimization histories as concrete examples of trustworthy AI usage in the laboratory setting. The node will also create open-access educational modules spanning experimental design, statistical learning, and AI-guided chemical discovery. A TRI-institutional architecture separates experimental execution, independent quality control/data curation, and AI model serving, providing built-in reproducibility and scalability probes. Further cross-node collaborations will unify metadata schemas, workflow templates, and transfer-learning strategies to non-chemistry domains where analogous multivariable systems may?benefit?from chemistry-derived descriptors and surrogate models. These contributions position the chemistry node as a scalable reference implementation for the national test bed, supporting interoperable autonomous experimentation across scientific domains. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the foundation's intellectual merit and broader impacts review criteria.- subawards are planned for this award.