Official company information and prototype website for 엘리트에듀 (Elite Edu), trading as Teacher Korea Consulting. Our original recruitment site remains available.

About

About Teacher Korea Consulting

We recruit English teachers for schools and academies in Korea. People run every part of the process. We are building software to help those people check job details more carefully, and we say plainly which parts exist today.

Facts

Legal trade name: 엘리트에듀 (Elite Edu). Teacher Korea Consulting is our recruiting brand.
Business opening date: 8 November 2024 (2024-11-08), confirmed by the business registration certificate.
This website: Supplementary official company information and prototype website. Our original recruitment website remains available, and recruiting applications continue through Gmail.
What we do: Human-reviewed recruiting that connects English (ESL) teachers with Korean schools.
Who we serve: Korean schools and academies hiring English teachers, and English teachers looking for work in Korea.
How to apply: By email to teacherkorea.consulting@gmail.com.
Product in development: Job Facts Check - finds missing or unclear facts in job descriptions before a person writes the bilingual draft. Product page.
Use of AI today: Claude Code was used to develop this site and our internal prototypes. No AI model is connected to the demo or makes decisions about applicants.
Decisions: Hiring is decided by schools and teachers. Visa and immigration decisions are made by the authorities. Our software does neither.

Stage: live, prototype, planned

Live Recruiting website and email applications. Our original recruitment site and the email application route above are in use.
Prototype Job Facts Check browser demo. Fixed rules checked against synthetic examples on the product page. No network, no storage, no AI model.
Prototype Offline job-facts validator. An internal Python tool that blocks drafts with missing job facts, unverified posting costs or unconfirmed channel permissions. It runs locally, uses no API and sends nothing.
Prototype Internal marketing tools. A content calendar and a photo-and-music video pipeline for our own social posts, run and reviewed by people.
Planned Claude API integration. Normalising school job wording into fixed fields, consistent English–Korean drafts and missing-field explanations. A recruiter reviews every output. Not built or verified yet.

How we treat data

Only what is needed

Job checks need school-confirmed job facts. Applicant information would only be used with that person's consent, for their application.

The demo keeps nothing

The demo runs in your browser. Its page blocks network connections and form submission, and it saves nothing.

Claude integration: a focused evaluation plan

The goal is clearer job information for teachers and less manual reconciliation for recruiters, not automated hiring. This is a development plan, not a claim of measured results or a deployed API integration.

1. Structured job facts

Evaluate Claude on synthetic Korean and English school messages to extract salary, pay period, schedule, location and start date into a fixed schema. Unknown facts must remain unknown, rather than being completed by the model.

2. Consistent bilingual drafts

Compare English and Korean drafts against the same source facts. Check that amounts, dates, hours and conditions survive translation unchanged. A recruiter approves every draft before sharing.

3. Evidence and missing details

Require each extracted fact to point back to its source text. Evaluate missed fields, unsupported additions and false warnings against human-labelled synthetic examples. Start with the local rule-based checker as a baseline.

4. Safe, practical rollout

Measure response time and estimated cost per reviewed job sheet during a future API pilot. Do not include real candidate personal data in the initial evaluation. The system must not score candidates or decide hiring, work eligibility or visas.

Support we would value from Anthropic

API credits for a bounded pilot, Applied AI guidance on model selection and evaluation design, structured-output reliability, Korean-English quality, privacy and safety, and latency/cost optimisation. Claude Team access, if eligible, would support evaluation of internal development and review workflows.

Success criteria and benchmark results will be published only after tests have actually been run. We do not claim customer counts, performance improvements, external funding or international subsidiaries that have not been verified.