Independent Research · Taiwan

Physical AI & Computational Biology

Cross-domain transfer learning for physical sensing systems, molecular docking pipelines, and AI-assisted drug discovery. Affiliated with Fu Jen Catholic University.

Clinical Neuroimaging Collaboration
IEEE Member · Power & Energy Society
NVIDIA DGX Spark GB10
Fu Jen Catholic University
Research Areas
Computational Drug Discovery

Computational Drug Discovery

MolGate — 27-stage QC pipeline for AI-generated docking poses and downstream safety checks. FemtoForge — quantum enzyme engineering for catalyst and binding-site optimization.

● Active — MolGate QC online; FemtoForge in integration
Clinical Neuroimaging

Clinical Neuroimaging

DTI fiber tract viewer for clinical tractography QA. TSM — topological semantic manifold on MDF distance matrices. Persistent homology analysis for bundle structure and asymmetry.

○ Ongoing — clinical cohort expansion
Physical AI · Materials

Physical AI & Materials

TDA-informed materials dataset (HF-CCD). Cross-domain transfer learning methodology for physical sensing systems across plasma, grid, and industrial monitors. Manifold causal models for interpretability.

● Active — HF-CCD under review · transfer prototypes
Publications
2026

TokaMind for Power Grid: Cross-Domain Transfer from Fusion Plasma In Preparation

TaiScience Research Group · Institute for Basic Science, Korea · First cross-domain validation of TokaMind outside nuclear fusion. F1 = 0.837 ± 0.040 on GESL/PNNL 500-event benchmark.

2026

A Curated High-Fidelity Carbide Materials Dataset (HF-CCD) Under Review

Data in Brief · Manuscript DIB-D-25-02507

2026

Manifold Causal Models Reveal Cryptic Drivers in Breast Cancer In Preparation

Targeting npj Precision Oncology · ASCL2 ranked #15 vs #15,000+, P=0.0048, validated in 6,253 patients.

Demo

3D Tokamak Plasma Visualization (Browser) Interactive

Real-time WebGL reconstruction of tokamak plasma equilibrium inspired by spherical-tokamak-style parameters: Shafranov-shifted D-shaped cross-section, magnetic field traces, orthogonal R–Z poloidal slice with “onion-ring” ρ–θ mapping, perturbation amplification along vertex normals for presentation (slider 10×–50×), and cockpit-style fused-state KPIs aligned with visualization notes on geometric exaggeration.

Live preview above — drag to orbit, scroll to zoom. Geometric deformation can be scaled in the full workspace.

Open full interactive workspace →
Demo

CSD-EWS: Pole & Damping Ratio Simulator Interactive

Educational demo mapping Jacobian eigenvalue proxies (σ ± jω) to damping ratio ζ, s-plane pole locations, and illustrative time-domain decay. Unified 5% ζ alert rule for weak-damping vs unstable (σ > 0) modes — aligned with power-system oscillation monitoring concepts (PMU / ModeMeter). Bilingual UI (中文 / English).

Open pole & damping simulator →
Infrastructure
Primary Compute
NVIDIA DGX Spark GB10 — 128 GB Unified Memory
Affiliation
Fu Jen Catholic University, Taiwan
Professional
IEEE Member · Power & Energy Society
Contact
steve.anderson@taiscience.org