Sam Blau
@samblau.bsky.social
Research scientist & computational chemist at Berkeley Lab using HT DFT workflows, machine learning, and reaction networks to model complex reactivity.
We're also releasing baseline models trained on OMol. To guide future MLIP development, we built novel evaluations on intermolecular interactions, conformers, and charge/spin. We hope to include frequency, ΔG, and TSopt tasks when we put up a public leaderboard in the summer 4/N
May 14, 2025 at 8:55 PM
We're also releasing baseline models trained on OMol. To guide future MLIP development, we built novel evaluations on intermolecular interactions, conformers, and charge/spin. We hope to include frequency, ΔG, and TSopt tasks when we put up a public leaderboard in the summer 4/N
OMol was constructed via an unprecedented diversity of methods: MD, ML-MD, RPMD, rattling, Architector, rxn path interpolation, AFIR, optimization, and scaled separation. We also recalculated some previous datasets and did additional sampling/structure generation atop others 3/N
May 14, 2025 at 8:54 PM
OMol was constructed via an unprecedented diversity of methods: MD, ML-MD, RPMD, rattling, Architector, rxn path interpolation, AFIR, optimization, and scaled separation. We also recalculated some previous datasets and did additional sampling/structure generation atop others 3/N
OMol covers 83 elements, a wide range of intra and intermolecular interactions, explicit solvation, reactive structures, conformers, charges -10 to 10, 0-10 unpaired electrons, and 2-350 atoms per snapshot. It required >6B CPU hrs, 10x more than any prev MLIP training dataset 2/N
May 14, 2025 at 8:53 PM
OMol covers 83 elements, a wide range of intra and intermolecular interactions, explicit solvation, reactive structures, conformers, charges -10 to 10, 0-10 unpaired electrons, and 2-350 atoms per snapshot. It required >6B CPU hrs, 10x more than any prev MLIP training dataset 2/N
The Open Molecules 2025 dataset is out! With >100M gold-standard ωB97M-V/def2-TZVPD calcs of biomolecules, electrolytes, metal complexes, and small molecules, OMol is by far the largest, most diverse, and highest quality molecular DFT dataset for training MLIPs ever made 1/N
May 14, 2025 at 8:52 PM
The Open Molecules 2025 dataset is out! With >100M gold-standard ωB97M-V/def2-TZVPD calcs of biomolecules, electrolytes, metal complexes, and small molecules, OMol is by far the largest, most diverse, and highest quality molecular DFT dataset for training MLIPs ever made 1/N
Looking forward to speaking at ACS on Sunday at 5:30! Come learn about "Popcornn" - a new method for double-ended transition state optimization atop machine learned interatomic potentials that is substantially better than NEB or GSM.
March 21, 2025 at 11:32 PM
Looking forward to speaking at ACS on Sunday at 5:30! Come learn about "Popcornn" - a new method for double-ended transition state optimization atop machine learned interatomic potentials that is substantially better than NEB or GSM.
Applications closing in one week! If you’re interested in a prestigious postdoc at the intersection of AI/ML and nuclear nonproliferation, don’t hesitate to apply - come work with me on fascinating f-block chemistry and computational/ML methods! (Must be a US citizen)
January 24, 2025 at 9:09 PM
Applications closing in one week! If you’re interested in a prestigious postdoc at the intersection of AI/ML and nuclear nonproliferation, don’t hesitate to apply - come work with me on fascinating f-block chemistry and computational/ML methods! (Must be a US citizen)
Example nanoparticle heterostructure optimization, driven by gradients of UV emission with respect to layer thicknesses and dopant concentrations from our hetero-GNN (not accessible from kMC) and sub-second inference (vs days from kMC) #F24MRS
December 2, 2024 at 3:52 PM
Example nanoparticle heterostructure optimization, driven by gradients of UV emission with respect to layer thicknesses and dopant concentrations from our hetero-GNN (not accessible from kMC) and sub-second inference (vs days from kMC) #F24MRS
Excited to speak at #F24MRS Thurs 1:30 - 1st talk of my career w/o any DFT connection - we design a hetero-GNN for learning core-shell nanoparticle properties, train on first ever large-scale NP kMC dataset, and use autodiff to optimize -> discover far OOD heterostructures with >6x enhanced emission
December 2, 2024 at 3:51 PM
Excited to speak at #F24MRS Thurs 1:30 - 1st talk of my career w/o any DFT connection - we design a hetero-GNN for learning core-shell nanoparticle properties, train on first ever large-scale NP kMC dataset, and use autodiff to optimize -> discover far OOD heterostructures with >6x enhanced emission