Google Scholar · LinkedIn · hpsahasrabuddhe@lbl.gov
I am a PhD candidate in Materials Science and Engineering at UC Berkeley and Lawrence Berkeley National Laboratory, in Anubhav Jain's group. I work on machine learning for materials at scale. I built the Materials Project's Harmonic Phonon Database, with the phonons of about 26,000 inorganic crystals from about 238,000 node-hours on NERSC Perlmutter and NREL Kestrel. I benchmark and fine-tune machine-learned interatomic potentials such as MACE for phonons and thermal expansion. Most of my code goes into atomate2 and the Materials Project.
- A High-Throughput ab initio Database of Harmonic Phonon Properties for Inorganic Crystals. H. Sahasrabuddhe et al. ChemRxiv (2026), under review at Scientific Data. doi:10.26434/chemrxiv.15004632/v1
- Atomate2: modular workflows for materials science. A. M. Ganose et al. Digital Discovery (2025). doi:10.1039/D5DD00019J
- Zatom-1: Towards a Multimodal Foundation Model for 3D Molecules and Materials. A. Morehead et al. arXiv (2026). arXiv:2602.22251
- AlabOS: a Python-based reconfigurable workflow management framework for autonomous laboratories. Y. Fei et al. Digital Discovery (2024). doi:10.1039/D4DD00129J
mlip-phonon-benchmarks computes phonons from 100 to 900 K with three MACE potentials and compares them with ab initio MD at 300 K. It also checks the Born charges and dielectric constants of MACE-Field against DFPT for 50 held-out materials.
The open and planned atomate2 PRs of our group are listed in this tracking issue.
| Pull request | Status |
|---|---|
| Lattice dynamics workflow using Pheasy | merged Sep 2026 |
| Harmonic lattice dynamics workflow using hiPhive | merged Sep 2026 |
| Thermal expansion workflow (CTEMaker) | merged Oct 2026 |
| Finite-temperature phonon workflow | in review, as a series of PRs |
| CALPHAD phase diagrams with ATAT sqs2tdb | in review |
| Debye-Waller factors from phonons | in review |
| Pull request | Status |
|---|---|
| Force field Born charges and dielectric tensors with MACE-Field (atomate2) | merged Oct 2026 |
| Anisotropic Debye-Waller factors in the XRD, neutron and TEM calculators (pymatgen) | in review |
| Seeded Langevin forces and a Nose-Hoover chain NVT preset for ASE MD (atomate2) | in review |
| JobStore document format as a pydantic model (jobflow) | merged Oct 2023 |
| Integration of Matbench Discovery (matbench) | merged Mar 2023 |
| "Go to page" navigation in the web GUI (FireWorks) | merged Mar 2026 |
| New mock decorator and config update (alabos) | merged Feb 2024 |
| Pull request | Status |
|---|---|
| Thermal displacement matrices on an odd q-point mesh, without the acoustic modes at Gamma (atomate2) | in review |
| Unpin pydantic after the emmet-core fix (atomate2) | in review |
| Converge the pheasy harmonic LASSO fit (atomate2) | merged Oct 2026 |
| Phonon thermal properties from phonopy's sum over the q-point mesh (atomate2) | merged Oct 2026 |
| Serialization with pydantic 2.14 (emmet) | merged Oct 2026 |
| Git dependencies moved to dependency groups, so that atomate2 can be released on PyPI (atomate2) | merged Oct 2026 |
| Fix pheasy anharmonic fitting and add fit options (atomate2) | merged Oct 2026 |
| Fix the acoustic sum rule in PhononBSDOSDoc (emmet) | merged May 2026 |
| Fix the Clarke thermal conductivity key between pymatgen and atomate2 (atomate2) | merged Mar 2026 |
| Pydantic v2 migration of atomate2 (#558, #565, #566, #567) | merged Oct 2023 |
| Use dumpfn as the default JSON writer (matbench) | merged Apr 2023 |
- atomate2: phonon database preprint in the pheasy docs, Zenodo DOI, ChemRxiv citation
- matbench: scaled error formula for classification tasks, link to Matbench Discovery on the leaderboard
- alabos: tooling to streamline the commit and PR flow, installation docs
