We help research groups and materials companies accelerate corrosion, catalysis, battery, coating and wide-bandgap semiconductor (SiC/GaN) development — from adsorption screening and MLIP training to ML multi-objective process optimization, delivered with peer-review-grade quality.
FOUNDER
Studied in the United States. Decades of experience in materials science. Founded Hainan Shuimu New Materials Co., Ltd., leading a professional computational team focused on computational materials services. Team track record: peer-reviewed SCI publications, an in-house inhibitor database, and a multi-system adsorption benchmark (in close agreement with DFT).
"We don't replace experiments — we filter out 90% of the failures before they happen, so experiments only test the most promising candidates."
Fixed-scope services with transparent pricing. Every deliverable includes raw data, publication-quality figures and a full methodology report.
Train or fine-tune DPA-4 / DPA-3.2 potentials for your target system (Fe, Cu, alloys, coatings). Model file + training report + validation (energy/force MAE, MD stability).
from USD 800 / projectBinding / adsorption energy screening of molecules on metal surfaces (fcc/bcc, multiple facets) with DPA or CHGNet — 50–300 systems per batch, ranked and benchmarked.
from USD 300 / batchNVT / NVE simulations of adsorbate stability, diffusion and interface behaviour (300–400 K). Trajectory analysis + energy statistics + stability ranking.
from USD 400 / systemDensity functional theory single points, geometry optimisation and adsorption energy validation to cross-check MLIP predictions — the dual-evidence standard.
from USD 250 / structureEnd-to-end support: calculations, figures, tables and methodology text for a manuscript — ready for submission to peer-reviewed journals.
from USD 1,200 / projectOne-on-one guidance on MLIP workflows, dataset preparation, DPA deployment on your cluster, or review of your simulation strategy.
USD 80 / hourSurrogate modeling (GP/XGB/NN) on small datasets + NSGA-II/Bayesian optimization for multi-objective parameter search — e.g. 4H-SiC CVD epitaxy process optimization. Cross-validation and uncertainty quantification included.
USD 1,200–2,200 / projectSiC (4H/6H/3C) / GaN band structures, defect chemistry, doping and electronic structure (PBE/HSE06), supporting epitaxy (CVD/MOCVD) related simulation.
from USD 300 / projectPublication-grade figures (300 dpi, journal-standard), 3D molecular/material rendering, data visualization (Pareto/heatmap/energy profiles) and academic slide layout.
from USD 50 / figureA simple, transparent 4-step process. You always know what you get and when.
Send your system details (structure files, target properties, deadline). We reply within 24 h with a fixed quote and timeline.
50% deposit to start. We run the simulation on our GPU cluster with rigorous input validation and QC checkpoints.
Every result passes three-gate quality control (energy magnitude, element consistency, requirement match). Deliverable package: data + figures + report.
Two rounds of free revisions. Final payment on delivery. NDA and IP protection available on request.
Not a reseller of GPU hours — a research-grade computational team with a real publication record.
peer-reviewed articles in computational corrosion and materials screening, with steady output across leading journals in the field. Your results meet journal standards by default.
Native DPA-4 / DPA-3.2 / CHGNet workflows with DeePMD-kit 3.2, the same family of models topping Matbench Discovery. Fine-tuning to your target chemistry available.
MLIP screening cross-checked with DFT (CP2K) validation — the standard that avoids the systematic-error criticism of pure ML predictions.
NDA by default. Your structures and data are never shared with third parties or reused for other clients. Clean-room processing on dedicated instances.
Not claims — numbers. Every service is backed by reproducible, cross-validated results.
Cross-metal, multi-facet inhibitor screening (17 metals × 3 facets × 7 molecules) run on DPA-2.4 in under 2 minutes per batch.
Recorded & documentedCross-backend validation: results in close agreement, with mean differences at the sub-meV scale. Reproducibility is our commitment.
Cross-validatedAn in-house database linking experimental efficiency with theoretical adsorption energies — the first of its kind.
In-house databaseFull package: raw data + input scripts + publication-quality figures + methodology report. Two rounds of free revisions; if a calculation fails to converge, we re-run at no cost or refund the affected part.
A client brings their own small DFT dataset — we adapt a general-purpose foundation potential (DPA-4) to their system in about a minute.
Force prediction accuracy improves markedly after fine-tuning on a small set of client DFT data, while generalization to unseen systems stays stable.
Full fine-tuning completes in ~1 minute on a single RTX 5090. Deliverable: training report + comparison figures + deployable model (19 MB) ready for ASE / LAMMPS / CP2K.
Your data, your system, publication-grade results. Bring your own DFT frames — we fine-tune, validate with MD stability tests, and hand you a ready-to-deploy model.
Typical turnaround is 3–7 working days depending on scope. Rush delivery (+50%) available for urgent projects.
Structure files (CIF, XYZ, POSCAR), the target surface/facet and the properties you need. If you don't have structures, we can build them for you.
Bank transfer or Payoneer/Wise. 50% deposit to start, 50% on delivery. Invoices in English or Chinese available.
Two rounds of free revisions. If the calculation itself fails to converge, we re-run at no cost or refund the affected part.
Yes. NDA by default, dedicated compute instances, and no reuse of client data across projects.
We sell methods, tools, data and compute — we do not provide ready-made product formulations or recipe optimisation for specific commercial products.
Yes — co-authorship and collaboration agreements are welcome, especially for computational + experimental teams.
Yes. We prepare data (DFT → DeePMD format), train/fine-tune DPA models on your data, and deliver the model + evaluation. Your dataset stays yours.
Send your system details and requirements — we reply with scope, price and timeline.
Hainan Shuimu New Materials Co., Ltd. · Haikou, China · Response within 24 h (CST)
Hainan Shuimu