Manufacturing AI / Data Science Engineer. 25 years on the factory floor.
I build the data science and AI that manufacturing runs on: yield prediction, anomaly detection, drift monitoring, and the smart-factory infrastructure underneath. Over the past two and a half years I independently delivered 30+ enterprise systems for a semiconductor materials manufacturer (MES, AI visual inspection with YOLO11, IoT automation, enterprise AI chatbot), all following ISO 27001:2022 principles.
Before that, I ran a software company for 19 years, delivering factory systems across 4 TSMC fabs and managing teams of up to 30 engineers. (Also an ordained Shingon Buddhist priest, a rare domain crossover.)
- Building data science demos end to end: Fab Analytics (BigQuery ETL, XGBoost / CatBoost yield prediction, Isolation Forest anomaly detection, PSI/KS drift, weekly retrain) and AMC Analytics
- Targeting Manufacturing AI / Data Science / FDE roles in Japan (Randstad / ExecutiveSearch.AI)
- Maintaining production demos (Shukuyodo: Nuxt 3 + Go) and 4 Go industrial repos (SECS/GEM driver, edge gateway, OT security scanner, shared API gateway)
| Data science across the line | Yield prediction, anomaly detection, drift monitoring, MLOps (XGBoost / CatBoost / Isolation Forest / BigQuery) |
| 30+ enterprise systems delivered solo (MES, quality, IoT, AI vision) | Following ISO 27001:2022 |
| 4 TSMC fabs, 19 years running a software company | Up to 30 engineers managed |
| Shingon priest + semiconductor developer | Rare domain crossover |
| 11 production languages | ZH (Native) / EN (Professional) / JA (JLPT N2, BJT J3) |
End-to-end semiconductor data science, on simulated fab data. Wafer / lot / equipment / sensor / defect schema across nodes N3-N14. Yield prediction (R² 0.80 / 0.81 ensemble), anomaly detection (F1 0.98), feature importance, AutoML comparison, drift dashboard. All data simulated, not derived from any company.
| XGBoost / CatBoost yield prediction | Isolation Forest anomaly detection | BigQuery ETL + lineage + MLOps (PSI/KS drift, weekly retrain) |
Next.js 16 + FastAPI · BigQuery + Vertex AI + GCP
25 years of factory knowledge, in one interactive system. Reads live factory data, flags NCRs with ranked options, and schedules orders. BCM simulation with cascading impacts and RTO/RPO matrix. 5 factory agents in a collaboration network.
| 54 CRUD APIs | 9 Production Stages | 15 AI Tools · 3 Languages |
Vue 3 + PrimeVue + TypeScript · FastAPI + SQLModel + Neon · Claude AI (Tool Use)
4 repos forming a complete industrial edge stack. All single Go binaries. Cross-compile to ARM64.
|
SECS/GEM driver — 12 SEMI standards, 5 protocols, IEC 62443 SL4, single binary |
Device bridge + Modbus scanner — plugin architecture, scan-to-config, ARM64 |
|
OT/ICS security scanner — CVE detection, IEC 62443 + NIST CSF 2.0 compliance |
Shared Go API gateway — Chi + pgx + JWT/Logto, embedded monitoring dashboard |
| Category | Projects |
|---|---|
| Enterprise AI & Manufacturing |
Smart Factory Demo — 30+ enterprise systems, MES, quality, IoT, AI vision |
| Language Learning |
ai-english-tutor — Voice-first speaking practice with AI grammar correction |
| Developer Tooling |
dash-devtools — Validation, E2E, AI vision CLI (on PyPI: pip install dash-devtools) |
| Area | Technologies |
|---|---|
| AI/LLM | Claude API (Tool Use), YOLO11, OpenCV |
| Enterprise | MES, Digital Transformation, Solution Architecture |
| SECS/GEM | HSMS, OPC-UA, MQTT, Modbus TCP |
| Security | ISO 27001:2022, IEC 62443, OWASP Top 10, AI Red Teaming |
| Frontend | React 19, Next.js 16, Vue 3, Nuxt 3, Angular 21, TypeScript, shadcn/ui, PrimeVue, PrimeNG |
| Backend | Go (Chi + pgx), FastAPI + SQLModel, Node.js |
| Database | PostgreSQL (Neon), Prisma ORM |
| IoT | Modbus TCP, OPC UA, RFID, WebSocket, SECS/GEM, Electronic Scale (RS-232) |
| OCR / PWA | glm-ocr (handwritten pig ID), Tablet-first PWA |
| Cloud | Vercel, Render, Neon, GitHub Actions |
Chinese (Native) / English (Professional) / Japanese (JLPT N2, BJT J3)
AI-assisted development with Claude Code.