Autonomous Analysis and Discovery

EMSL's Autonomous Analysis and Discovery (AAD) Integrated Research Platform (IRP) positions our facility to adopt, adapt, create, and deploy AI agents, agent-enabled workflows, and integrated computational infrastructure that support the full research cycle. These efforts span hypothesis generation, experimental and computational study design, data analysis and integration, model development, and the planning of new laboratory experiments and modeling studies.

The AAD IRP advances the AI-agent systems needed to enable automated and autonomous science across EMSL's experimental platforms, modeling expertise, data resources, and scientific workflows. These systems are fundamental to autonomous search and retrieval, agent ecosystem architecture, workflow orchestration, model-experiment integration, human-in-the-loop decision-making, and automated knowledge discovery. Together, they help bridge gaps across the design-build-test-learn cycle by enabling researchers to move more effectively between data, models, experiments, and scientific interpretation.

Within EMSL, the AAD IRP serves as a coordinating and enabling layer for AI-agent adoption and development. It does not replace the deep scientific, computational, or experimental expertise housed in EMSL's other IRPs. Instead, AAD works across those IRPs to identify where AI agents, workflow automation, model integration, and emerging infrastructure can increase scientific impact. This includes close alignment with the Data Transformations IRP on data curation, workflow development, statistical analysis, visualization, and computational automation, as well as with the Systems Modeling IRP on AI-enabled modeling, model-experiment integration, simulation workflows, and computational approaches that support predictive understanding of biological and environmental systems. AAD is grounded in EMSL's scientific computing, data, and experimental ecosystem, connecting AI-agent workflows to resources such as Tahoma, EMSL's data repository housed in the Aurora archive, scientific software, visualization environments, and instrument-facing workflows. A central element of this ecosystem is BRIDGE, a Department of Energy (DOE) wide curated data lakehouse that will enable access to EMSL and DOE Office of Science's Biological and Environmental Research (BER) program data, instrument outputs, and lakehouse-connected resources for AI-agent workflows.

The AAD IRP also supports EMSL's Modeling, Integration and Data Agents for Science (MIDAS) strategic science objective by establishing the technical expertise, leadership, and operational practices needed for sustained deployment of a large number of AI agents specialized for different data types crucial to advancing BER-relevant science. This includes connecting large language models, scientific software, hardware interfaces, data transformation pipelines, and human oversight into reliable workflows for scientific discovery.

Specific areas of AAD emphasis include:

• Developing AI-agent architectures and workflows to accelerate BER-priority science across EMSL's experimental, computational, and modeling resources.  
   
• Advancing agent-enabled infrastructure for autonomous search, retrieval, analysis, reasoning, and knowledge discovery.  
   
• Supporting integration between AI agents, scientific software, data repositories, visualization environments, EMSL computing resources, and instrument-facing workflows.   
   
• Collaborating with EMSL's Computing and Data Operations group on infrastructure, governance, policy, security, data access, and responsible use of AI-agent systems.  
   
• Working with EMSL Science Area Leads and IRP leads (experimental IRP leads as well as the IRP leads for CAM's other IRPs, Data Transformations, and Systems Modeling) to connect AI agents with laboratory systems, computational workflows, and data transformation pipelines.  
   
• Supporting model-experiment integration (BER's ModEx paradigm) by helping researchers move between experimental observations, computational models, simulation outputs, and new experimental or modeling studies. 

Building on this foundation, the AAD IRP advances AI-agent systems that are tightly coupled to EMSL's flagship experimental platforms, modeling and computing resources, DOE Genesis Mission goals, BER data resources, and Facilities Integrating Collaborations for User Science (FICUS) affiliated systems. These efforts enable more effective autonomous experimentation across EMSL and the broader DOE complex, increase the scientific return of advanced analytical and computational workflows, and operationalize automated knowledge discovery in support of DOE Office of Science missions and goals.