02.08.2026 aktualisiert
100 % verfügbar
Data & AI-Automation Engineer | Python ETL, Scraping, Korean Data & OCR QA
Siheung-si, Südkorea
Weltweit
Self-taught: Python, data engineering, applied LLM work. Public code at jigonyoo.comÜber mich
Data and AI-automation engineer based near Seoul. Korean is my first language, so Korean documents, forms and OCR output are things I audit natively rather than guess at. I work fully async, in writing. What I care about: pipelines that tell you when they are wrong, not only when they ran.
Skills
APIsSoftware für Optische ZeichenerkennungWeb ScrapingETLData TransformationData MiningDatenqualitätPythonMachine LearningSQLitepandasData ScrubbingDaten-Pipeline
Data and AI-automation engineer working in Python. I build pipelines that pull data from APIs, databases and messy public sources, clean and structure it, load it into Postgres or a dashboard, and then put the whole thing on a schedule so it keeps running unattended.
Core stack: Python, pandas, NumPy, SQL, PostgreSQL, Supabase, SQLite, REST APIs, JSON, CSV, Excel, Docker, Git, GitHub Actions, AWS Textract, Next.js, React, Vercel.
Areas I work in:
ETL and data pipelines - scheduled extraction, normalisation, idempotent loads, schema migrations, row-level security, freshness monitoring.
Web scraping and data mining - public sources, retry and backoff, change detection, structured output to CSV or Excel.
Document AI - invoice and PDF extraction, OCR, field-level confidence scores, reconciliation against source totals.
Korean-language data - Korean is my first language. I audit Korean documents, forms and OCR output natively rather than guessing at them: spacing around particles and bound nouns, NFC/NFD normalisation of Hangul jamo, and mixed Korean-English strings that break naive tokenisers. Also Korean-language scraping targets and Korean/English bilingual QA.
LLM and RAG - citation-grounded question answering that refuses uncertain answers, evaluation harnesses, regression tests for prompt pipelines, retry, timeout and cost guards for agent tools.
Analytics and machine learning - churn prediction, feature pipelines, model cards, leakage checks, scheduled reporting.
Dashboards - interactive boards with filtering, sorting and daily auto-refresh.
How I work: verification is part of the build rather than an afterthought - confidence scores, checks against the source, and a human-review queue for anything uncertain. Every project ships with tests and a plain-English handover document, so the client never has to touch the code.
Currently in production for a client: a daily sports-prediction pipeline - scrapers, model outputs, a normalised Postgres schema with row-level security, and a scheduled job that refreshes the board every morning.
I work project-based at a fixed price rather than hourly, so the cost is known before the work starts. I am early in my freelance career and would rather point at what is running than claim a number of years.
Public code and live examples: jigonyoo.com / github.com/jigonyoo
Erfahrung
Berufserfahrung0 - 5 Jahre
Freelancer seit2026
Sprachen
EnglishgutKoreanMuttersprache
Projekthistorie
Built and maintained the data layer for a US sports-analytics site (ongoing, three consecutive contracts).
Delivered a WNBA player-prop model end to end: scraping the source data, generating projections, and publishing a live board. Migrated the site's storage from in-repo files to Supabase (Postgres) while preserving its existing daily schedule.
Audited an inherited scoring pipeline and found a defect that was silently producing zeros: one line selected a DataFrame column by position and landed on the wrong one after an upstream change, so the join produced nulls that were coerced to 0. Nothing raised an error, and the dashboard displayed 0 as a valid score. Traced it end to end, wrote it up with severity, and delivered a plain-English handover the client's next developer could act on without me.
Stack: Python, pandas, FastAPI, PostgreSQL/Supabase, GitHub Actions scheduled runs, React.