Optics Ph.D. with 8+ years in optical & semiconductor metrology (SEM, AFM, Raman, NSOM), building AI applications that don't just predict — they report how much you can trust the prediction, with GUM-compliant uncertainty budgets (JCGM 100/101).
- 🔬 Measurement scientist first: ISO 18516 round-robin lead contributor · KOLAS-accredited calibration & testing background
- 📐 I build the full loop: physics forward model → inverse solver → GUM uncertainty budget
- 🧰 Domains: OCD / scatterometry, XRR, SEM·TEM image analysis, AFM probe health, Raman/SERS QC
- 🤖 MCP builder: author of the first Model Context Protocol server for GUM measurement uncertainty
| Project | What it does | Stack |
|---|---|---|
| metrology-inverse | Forward → inverse → GUM uncertainty across 3 instruments: OCD (RCWA), XRR (Parratt), autodiff CD fitting — validated on real NIST scatterometry data (L100P300, 9 dies). Exact-Jacobian sensitivity for the uncertainty budget. | Python, PyTorch, Meent, refnx |
| measurement-uncertainty-mcp | First MCP server for GUM-compliant uncertainty analysis: Type A/B, Welch–Satterthwaite ν_eff, expanded U(k), JCGM 101 Monte Carlo, KOLAS-ready budgets. | Python, MCP |
| tiphealth | Recipe-aware HAR AFM tip predictive maintenance — hybrid Archard physics + ML, 8 industrial probes × 6 materials, NIST-calibrated wear rates, conformal prediction intervals, Dockerized API. | Python, Docker, Streamlit |
| spectraguard | Uncertainty-aware spectral QC for SERS/Raman/IR — 6-metric confidence score with bootstrap CIs, cross-instrument transfer, streaming SPC, CLI + CI pipeline. | Python, NumPy, SciPy |
| semiconductor-defect-classifier | Defect classification on SECOM fab sensor data (1,567 wafers × 590 sensors, 6.6% defect rate) — imbalance handling, Optuna tuning, honest K-fold evaluation. | Python, XGBoost, Optuna |
| semiconductor-ai-portfolio | Analysis pipelines on my own PhD measurement data: MoSe₂ photoluminescence peak/FWHM analysis, NSOM defect mapping, TMD comparison. | Jupyter, pandas, SciPy |
Model Context Protocol servers I built and maintain:
- measurement-uncertainty-mcp — 10 GUM/JCGM tools for ISO/IEC 17025 calibration labs, from any MCP client
- vertical-mcp org — Korean government open-data connectors: dart-mcp (corporate disclosures, OpenDART) · kolas-mcp (ISO/IEC 17025 accredited-lab registry) · ntis-mcp (national R&D projects & funding) · grant-mcp (NSF / ERC / NRF grant search)
- schedule-optimizer-mcp · notion-workspace-automation-mcp — productivity MCP servers
Metrology & standards: GUM (JCGM 100:2008) · JCGM 101 Monte Carlo · ISO/IEC 17025 · RCWA · XRR · SEM/TEM · AFM · Raman/SERS · NSOM