News and activities

Latest updates from the MLAMD Center

Recent research directions, meetings, publications, software releases, datasets, and project news from the MLAMD Center.

News and activities

Latest updates and project activities

Next

Agentic workflow layer

A supervised AI controller will help choose scientific next steps and manage HPC execution decisions such as task batching, retries, queue monitoring, CPU/GPU allocation, and provenance tracking.

May 2026

Beyond-ternary discovery pipeline

The current Thrust 1 effort extends the workflow from demonstrated ternary searches to multicomponent motif libraries, MLIP relaxation, DFT validation, and disorder-aware search.

Mar. 15-20, 2026

APS Global Physics Summit presentation

The project reached the APS Global Physics Summit program, extending visibility for MLAMD progress across the condensed-matter and materials-physics community. Session listing

Mar. 2026

exa-AMD Module 2 milestone

Thrust 1 advanced the next exa-AMD module toward MLIP-assisted relaxation, adaptive search, and higher-order composition-space exploration.

Feb. 2026

exa-PD phase-diagram workflow

Thrust 2 has a working Module I story for massively parallel MD-based free-energy calculations, phase-diagram construction, and CALPHAD-compatible thermodynamic output.

Late 2025

exa-AMD released and published

The project released exa-AMD as open-source software and published the JOSS software paper, establishing a citable workflow foundation for community use.

Aug. 11-12, 2025

DOE BES PI Meeting

The center shared progress through the 2025 EFRC-Hub-CMS-CCS PI Meeting, connecting exa-AMD and exa-PD developments with the broader BES computational materials science community. Program book

June 4, 2025

GPN Annual Meeting work update

Ames National Laboratory presented CMS Center work on accelerating functional-materials discovery using AI/ML and exascale computing, with poster talks and handouts highlighting Thrusts 1 and 2. Meeting page

exa-AMD scaling plot showing near-linear performance as GPUs increase
Benchmarking from CMS materials shows strong scaling behavior for exa-AMD workflows.
Calculated Cu-Zr phase diagram from exa-PD workflow
exa-PD demonstration output: calculated Cu-Zr phase behavior from MD and thermodynamic modeling.