Hossein Tahmasbi

I am a computational materials scientist working at the intersection of machine learning, high-throughput simulation, and materials discovery. I am currently open to Materials Informatics and Applied ML-for-Science roles.

My research focus: building, and just as importantly evaluating, machine learning interatomic potentials (MLIPs) — deciding which models to trust, where they fail, and what the metrics actually mean. I combine AI/ML, density functional theory (DFT), and high-performance computing to accelerate materials discovery and connect predictions to physical reality.

• Benchmarking & validation — Led the benchmarking of leading universal MLIPs across elemental systems (MACE, MatterSim, PET), probing stability, transferability, and failure modes during global structure searches.
• High-throughput MLIP development — Trained neural-network interatomic potentials for aluminium, iron hydrides, and water using MongoDB/AiiDA-based automated workflows and custom PyTorch models.
• From predictions to physics — Validated large-scale neural-potential simulations against experimental IR spectra; discovered novel structures of TiO₂, ZnO, and near-stoichiometric MgO.

I hold a Ph.D. in Condensed-Matter Physics from the Institute for Advanced Studies in Basic Sciences (IASBS), with a visiting research stay at Leiden University, and currently work as a scientific software researcher at CASUS–HZDR in Görlitz, Germany.

Here you can find my research, CV, and publications on Google Scholar.

Contact: h.tahmasb@gmail.com | Google Scholar | GitHub | GitLab | LinkedIn