I run F&B brands at UNICUP COMPANY, plan projects with companies and public institutions, and build AI tools for day-to-day business work. I provide AI training for small businesses and organizations. I also build Python tools for data validation, CLI testing, and CI evidence, and contribute reproducible fixes to open-source projects.
F&B 브랜드를 운영하며 기업·기관 협업 프로젝트를 기획하고, 업무에 쓰는 AI 도구와 활용 교육을 제공합니다. 데이터 오류와 테스트 결과를 직접 확인할 수 있는 도구도 만듭니다. 한국어·영어로 질문과 제안을 받습니다.
Contact: ceo@unicupcompany.com
Career & public work / 경력과 활동 · AI training for organizations / 기관·기업 AI 교육
게재 승인 · Accepted article
Operational domain histories and later restaurant inspection citations in New York City — NHI Hospitality International Journal, October 2026.
My research combines foodservice operations with open data, statistical simulation, and reproducible software. Research repositories include source traces, executable analyses, and numerical results.
| Research project | What the repository contains |
|---|---|
| Bootstrap confidence-interval replication · Preprint | Seven interval methods, small-sample coverage experiments, and checks against Hesterberg's numerical targets. |
| Shared-ingredient cafe replenishment · Preprint | Generated-demand experiments comparing three replenishment policies across 54 settings, with separate calibration and evaluation runs. |
| Robert's truncated-normal algorithms | Independent calculations of 27 printed acceptance probabilities, an algorithm-switching boundary, and Monte Carlo checks. |
학술 심사 · Peer review — Completed one manuscript review for NHI Hospitality International Journal, October 2026.
저서 · Book — AI 프랜차이즈 혁명: 기술을 통해 경영을 바꾸다 (퍼플, 2025). ISBN 9788924174588
| Tool | Use it when | Run and install |
|---|---|---|
| JSON Repr Probe | Your CLI should return the same JSON result despite key order, whitespace, line endings, or Unicode escapes. Select a result subtree with --compare-pointer /data to leave changing metadata out of comparison. |
Examples · Wheel 0.2.0 · Container |
| JUnit Evidence Gate | A green CI report must contain actual testcase records and meet execution and skip limits. Emit GitHub error annotations with --format github. |
Examples · Wheel 0.2.1 · Container |
| Public Data Sentinel | CSV, TSV, or JSON data must satisfy explicit rules before entering a report. Preserve text identifiers such as 00123 and display field names literally in Markdown. |
Examples · Wheel 0.2.1 · Container |
These tools run offline with Python 3.10+ and have no runtime dependencies. Each repository includes installation instructions, runnable examples, and Windows/Linux CI.
git clone https://github.com/qorud02/json-repr-probe.git
cd json-repr-probe
python -m json_repr_probe --input examples/input.json -- python -m examples.stable_cliThe stable CLI passes. Replace examples.stable_cli with examples.order_sensitive_cli to reproduce a key-order bug and see the first changed JSON Pointer.
Calculate menu contribution after ingredients, packaging, and platform fees using Decimal arithmetic. It handles Korean CSV exports, VAT, negative margins, and optional fixed-cost scenarios.
Store and delivery comparison · Wheel 0.1.0 · Container · Calculation basis · Tests
- mediation #81: return two-sided analytic TSLS p-values.
- USGS dataretrieval-python #431: reject missing nearest-observation timestamps; included in v1.4.0.
- scikit-hep-testdata #240: isolate the cache failure test from network access and shared state.
- HDX Python API #114: correct resource-generation date examples to use explicit keyword arguments.
- kiwipiepy #236: support empty Kiwi format strings.
- ForestCI #128, #130, #132: correct bias correction, single-sample variance, and calibration with supplied sampling counts.
- ForestCI #134: release unselected class-probability arrays before stacking.
Start with a runnable example or a reproducible bug. The contributor guides explain the tests and review process:
JSON Repr Probe · JUnit Evidence Gate · Public Data Sentinel · F&B Menu Margin Kit