Every thread feeds the next. First-principles calculations generate data, machine-learned models compress it, agents search it, and our experimental collaborators close the loop.
DFT for formation energies, migration barriers, and the reference forces that train everything downstream.
MLIP-driven molecular dynamics at near-DFT accuracy, millions of steps, active learning in the loop.
Interfaces, grain boundaries, and degradation fronts — where real materials actually fail.
Continuum and FEM surrogates for cells, joints, and processes that engineers can build.
Multimodal representations carrying composition, structure and process in one embedding, so a model trained on one property transfers to the next — and generative models can run the design in reverse.
LLM-based agents and NLP pipelines that mine literature, build databases, plan simulation campaigns, and decide what to compute next — closing the discovery loop without a human at every step.
Domain-specialized and data-efficient MLIPs — fine-tuned for specific chemistries, validated on relaxation and interface adhesion, with uncertainty-driven active learning that calls DFT only where the model is wrong.
Bridging atomistic results to engineering scale: cohesive-zone models for multi-material interfaces, thermal surrogates for manufacturing, and safety models for cells under abuse.
High-throughput screening of cathodes and solid electrolytes across Li, Na, Ca and Zn chemistries, plus atomistic models of the cathode/electrolyte interfaces that govern lifetime.
ALD and organometallic precursors, thin-film interfaces such as TaN–Cu, Ru–SiO₂ and W–TiN, 2D heterostructures and MXene / MAX-phase chemistries — the same simulate-and-learn loop applied to device fabrication.
From zeolite frameworks to kirigami metamaterials — connecting atomistic stiffness to printed macro-structures, and searching design spaces too large to enumerate.
Kyoungmin Min has led the Computational Science and AI Lab since 2019, now at Yonsei University. His research joins first-principles simulation with machine learning to accelerate the discovery of energy and semiconductor materials.