A company for work that can’t be wrong.

Who we are

Sizlon Inc. (주식회사 시즐론) was founded by Kwangyeon Lee, a data-mining engineer who spent fifteen years building news and media data pipelines — collecting, cleaning, and structuring data that grows every year. The products are built and operated in-house: no outsourcing, no reselling.

Why we exist

Out of that data grew the AI era, and AI-generated answers are convenient and usually accurate. But fifteen years of handling data taught one thing: parts of enterprise work do not want usually accurate — they demand definitively accurate. A public tender where one missing form disqualifies the bid. A data pipeline that quietly returns the wrong number after a page changed shape. In that work, "mostly right" is not an answer.

How we build

One discipline runs through every product: AI proposes, a deterministic layer verifies, and only what holds is promoted. Every result carries a confidence score; low-confidence results are isolated for a human instead of shipped silently; every decision traces back to the evidence behind it. Nothing irreversible happens without a verified backup and an audit trail.

What we make

Two products today. Miriboa checks a 나라장터 bid package against what the notice actually requires, before submission — operated as a service at miriboa.sizlon.io. Crawler Platform runs web extraction that repairs itself instead of breaking silently — licensed and installed in your own infrastructure.

Both were designed from evidence rather than intuition: Miriboa’s checks come from 4.5 million rows of public bid-award records — the actual distribution of why bidders get disqualified.