Need a quick overview of the upcoming #F22MRS sessions? Check out anjohan.pythonanywhere.com built by MIRer Anders Johansson

Our MIRer Anders Johansson Anders Johansson talked about "Billions of Atoms with Machine Learning Interatomic Potentials—Performance Portability of FLARE" at #f22mrs. Check the paper below!

Our Allegro work is now published in Nature Communications 🎉 Allegro is as a step towards simulations that are truly predictive of experiment - a new combination of scale, speed, and accuracy. Paper: nature.com/articles/s4146… Blog: go.nature.com/3wPcSPw Code: github.com/mir-group/alle…

Allegro was selected as an Editor's Highlight in Nature Communications: nature.com/collections/wt…

I am presenting today at the APS March Meeting! #APSMarch

Packed rooms at our talks today at #APSMarch !

Upcoming talks from our group at #APSMarch today (Wed) 👇!

I am presenting my poster today at #APSMarch ! Come see how you can quickly compute thermal conductivities with either Green-Kubo or Boltzmann transport using FLARE (github.com/mir-group/flare) and Phoebe (github.com/mir-group/phoe…) I will be there from ~1.30

We are thrilled that Allegro was selected as one of the finalists for this year's "Nobel Prize for Supercomputing", the ACM Gordon Bell Prize. Congratulations to the team, pictured here happily finalizing the submission. arxiv.org/abs/2304.10061

Tomorrow I will be presenting in the materials&chemistry track at #SciPy2023 ! Allegro and FLARE are two very different machine learning models for how atoms interact. How do they work, how are they implemented, how do we make them fast, and which should you choose?

Congrats to Albert Musaelian, Anders Johansson, & Simon Batzner, #GordonBellPrize finalists for scaling the Allegro model for massive biomechanical systems—another exciting project powered by #Perlmutter! ow.ly/Pgbh50PA1HA Berkeley Lab Computing Sciences Area Berkeley Lab DOE Office of Science

FLARE with active learning and GPU-acceleration applied to direct simulations of surface reconstruction by Cameron. Explicit, visual comparisons with experiment are cool!

Running MD simulations with ML force fields? Consider learning the scale separation for a potential ~2-4x speed boost using Multi-scale integration: working paper: arxiv.org/abs/2310.13756 Great collaborating with Alby Musaelian, Anders Johansson, Tommi Jaakkola, and Boris Kozinsky

Alby and I are presenting at #SC23 tomorrow on Allegro and how to scale equivariant neural networks to thousands of GPUs on #HPC machines, with benchmarks on large biomolecules! Paper: dl.acm.org/doi/10.1145/35… Sneak preview:

Fantastic presentation on Allegro to an overfilled hall at the ACM Gordon Bell Prize session at SuperComputing ‘23. Monumental achievement by PhD students Alby Musaelian, Anders Johansson & Simon Batzner, who were nominated for the top prize in the field of scientific computing!

Happy to announce our Allegro model for unified learning of electric enthalpy, forces, polarization, Born charges, and polarizability through exact physical constraints. arxiv.org/abs/2403.17207 Boris Kozinsky Chuin Wei Tan Anders Johansson Cameron J. Owen Materials Intelligence Research @ Harvard Harvard SEAS

Excited to announce a new version of our arXiv paper "Unified Differentiable Learning of Electric Response", now including ferroelectrics in addition to dielectrics! arxiv.org/abs/2403.17207 Anders Johansson Chuin Wei Tan Cameron J. Owen (cr. video) Boris Kozinsky Materials Intelligence Research @ Harvard Harvard SEAS

Last month, we released a major update to the NequIP framework that fully leverages PyTorch 2.0 compilation for MLIPs. It’s significantly faster, easier to use, and more versatile than before. Preprint: arxiv.org/abs/2504.16068 Code: github.com/mir-group/nequ… nequip.readthedocs.io