The Transformational AI Models Consortium: ModCon

We establish the Transformational AI Models Consortium (ModCon), led by Argonne with major roles for Pacific Northwest, Berkeley, Oak Ridge, Jefferson, SLAC, NREL, and Idaho National Laboratories, along with supporting roles from Brookhaven, Ames, Savannah River, and Fermi National Laboratories in coordination with NNSA labs. ModCon provides the intellectual, organizational, and technical framework for building advanced scientific AI models, centered on the formation and support of Model Teams (MTs) that DOE selects to address mission-critical challenges. ModCon enables MT success by supplying access to baseline AI capabilities, standardized data preparation pipelines, workflow best practices, and partnership structures, while capturing MT requirements needed to drive the development of the American Science Cloud (AmSC) and ensuring models are delivered back to the community through AmSC. In doing so, ModCon fulfills the obligations of the OBBB Act (P.L. 119-21, Sec. 50404), which mandates coordinated AI models and infrastructure to accelerate U.S. discovery and innovation.

ModCon supports the process of MT selection through workshops, evaluation criteria, and proposal analyses. Once MTs are established, ModCon connects its crosscutting services into their work. ModCon does not duplicate domain-oriented modeling but ensures teams have access to AI-ready FAIR data, innovative evaluation practices, scalable agentic workflows, and secure partnership agreements. In turn, MTs define requirements for data, infrastructure, and APIs, shaping AmSC development. MTs also deliver predictive models with state-of-the-art reasoning that can analyze simulated and experimental data and, in some cases, enable active experimental control for automated laboratories.

ModCon’s work organizes into four crosscutting teams. Intellectual Property and Partnership Formation (IPFP) manages agreements, data rights, and structured industry/academic engagement. Data Brokering and Standards (DBS) builds standardized raw-to-AI-ready (RA→AIR) pipelines, enforces FAIR practices, and operates a “Data Detective” network linking MTs to DOE data sources. Best Practices in Scientific Workflows (BPSW) benchmarks AI-augmented workflows, establishes reproducible evaluation protocols, and produces training material and flexible templates to ensure measurable productivity gains. Baseline AI Capabilities (BASE) develops shared AI tools: multimodal reasoning front ends, agent-based data pipelines, evaluation harnesses, self-improving frameworks, safety/security protocols, and a jointly developed (with AmSC) core agentic framework to orchestrate workflows across HPC and cloud platforms. Together, these reduce duplication and accelerate MTs' adoption of AmSC services and APIs while prioritizing open source.

ModCon is led by a Director (Stevens, Argonne), a Deputy (Corley, Pacific Northwest), and an Executive Council of team leads. Each MT operates under a lightweight charter defining scope, milestones, datasets, evaluation protocols, and AmSC integration points. Engagement is maintained through biweekly standups, monthly demos, and quarterly reviews. Coordination with AmSC is formalized through liaisons, while external partners engage via structured touchpoints, annual meetings, and workshops.

Success is measured by MTs delivering functioning AI models aligned with DOE challenge problems, adoption of DOE data and workflow standards, AmSC integration, and evidence of accelerated outcomes. Metrics combine quantitative benchmarks (accuracy, robustness, time-to-solution) with case studies of mission acceleration. Sustained MT engagement and partner participation serve as leading indicators. ModCon succeeds only if MTs succeed—through coherent data, workflows, partnerships, and infrastructure.

In summary, ModCon establishes a durable consortium structure that supports MT selection and success while driving AmSC infrastructure requirements. By coupling DOE’s experimental and computational assets with frontier AI techniques, ModCon provides a reproducible, scalable framework for scientific models that accelerates discovery, delivers productivity gains, and reinforces U.S. leadership in science and technology.