Grade handwritten Korean from one photo.

Snap a handwritten answer. GLSSY scores the essay by TOPIK-style rubric, evaluates the handwriting itself, and returns red-pen corrections in seconds. No tablets, no typing.

For universities, language institutes & academies teaching Korean · TOPIK-ready

glssy.ai · grading
GLSSY teacher grading screen — handwritten Korean essay with AI red-pen corrections and rubric scoring
Live working service · TRL 7
0.85 QWK on a public benchmark
200K expert-labeled syllables
3 core patents filed
₩1B annual revenue ('25)
Handwriting featured in Netflix · EBS · tvN Calligraphy training for Samsung · Lotte · Hyundai

One teacher. 186 students. Every essay graded by hand.

Writing is the only part of TOPIK graded entirely by humans, so detailed feedback is the first thing that gets cut. Students queue for feedback; teachers drown in grading.

550K+
TOPIK test-takers per year, an all-time high.
186
students per King Sejong Institute teacher (216K learners, 1,163 teachers worldwide).
25K
Korean-language majors in Vietnam alone, across 60+ universities, chronic grading overload.

The only engine that grades real pen-on-paper Korean.

Automating this means solving handwriting recognition, rubric scoring and penmanship evaluation at once, which is why no one else has.

01

One photo, no hardware

A student photographs a real pen-on-paper answer with a phone. A phone camera is the only hardware required.

02

Dual-engine grading

One engine scores content, structure and language by TOPIK-style rubric. A second, patented engine evaluates the handwriting itself: spacing, alignment and stroke stability, pixel by pixel.

03

Teachers stay in control

AI does the first pass; your faculty confirms with one click. Every confirmed grade makes the engine better for your institution.

This is exactly what your faculty and students see.

These are our live grading screens, not a concept. Photograph an essay, and both sides get rubric scores and red-pen corrections in seconds.

What the student receivesScore, criterion breakdown & every correction explained
GLSSY student result screen with score and corrections
Admin & program dashboardsClasses, teachers and grading volume at a glance
GLSSY institution admin dashboard

Agreement with expert graders: 0.85

QWK (quadratic weighted kappa) measures how closely AI scores match a human expert, where 1.0 is identical. Measured on 83 public Korean-learner essays, double-graded by AI and experts. This year's target: 0.90.

Generic LLM (single prompt)0.28
GLSSY hybrid engine0.85
Human vs. human (upper bound)0.98
00.250.500.751.0

Why it works: expert anchor answers + handwriting features + rubric decomposition: an architecture, not a prompt. Your faculty stays the final authority.

Two people feel it immediately.

      What could GLSSY save your program?

      400
      4
      12
      Grading hours saved / month
      0 hrs
      Labor cost saved / year
      $0
      Teacher-days handed back / year
      0 days
      Illustrative estimate. GLSSY grades a first pass in about 60 seconds per essay (measured); your teachers review and confirm, so real savings depend on your workflow. See the product → then start a free pilot →

      See how institutions like yours use GLSSY.

      Filter by the kind of program you run.

      * Figures shown are from active 2026 pilots and internal benchmarks. Named case studies publish as each pilot completes.

      Frequently asked questions

      Written so both you and the AI assistants your colleagues ask get a clear answer.

      How accurate is GLSSY's grading?

      On a public Korean-learner writing benchmark it reaches QWK 0.85 against expert graders, versus 0.28 for a generic single-prompt LLM, and 0.98 for two human experts, measured on 83 double-graded essays. These are public-benchmark figures, not official TOPIK scores, and a teacher always confirms the final grade.

      Does it really grade handwriting from a photo?

      Yes. Students photograph a real pen-on-paper answer with a phone. A patented engine analyzes spacing, alignment and stroke stability at the pixel level, using nothing but a phone camera.

      Is it built for TOPIK?

      Yes. GLSSY scores the writing dimensions TOPIK assesses: content & task, structure, language range and accuracy. Practice maps directly to the exam.

      Can we use our own rubric?

      Yes. It grades against your criteria and level standards, and every grade your faculty confirms further tunes the engine to your program.

      What does it cost to try?

      The pilot is 100% free. We cover the app, grading, materials and support. It's non-binding, non-exclusive and starts with a single class. Just leave your email and we'll set you up.

      Is student data safe?

      Yes. It's web-based with no IT setup, works on any phone or PC, and is privacy-first: student consent, anonymized, and deletable on request.

      Korea's handwriting specialists — now teaching AI to grade it.

      For a decade, GLSSY LABS (글씨를 수놓다) has defined what good Korean handwriting looks like as Korea's leading handwriting-education brand. That expertise is the ground truth our AI learns from.

      Our handwriting has featured in Netflix, EBS and tvN productions, and our founders have taught calligraphy at Samsung, Lotte and Hyundai.

      10 yrs
      as Korea's only handwriting-education brand
      100K
      community across our store, KakaoTalk and social
      200K
      expert-labeled handwriting syllables, our AI's ground truth

      Leadership

      CEO · Business

      Wonjin Choi

      Grand Prize at the national Han Seok-bong calligraphy competition. Author of two handwriting workbooks and creator of Korea's first AI-handwriting contest. Pen-work for Netflix and EBS productions.

      VP · R&D

      Soyeon Park

      Calligraphy master with 20+ years who set the marking standards for our AI. Bestselling workbook author; corporate instructor for Samsung, Lotte and Hyundai; handwriting for a tvN drama.

      Behind the product: a dedicated AI engineering team (handwriting Vision + LLM) and a security partner recognized in KISA's White-Hat Hall of Fame.

      ₩1B
      Annual revenue ('25), 5 profitable years
      100K+
      Handwriting textbooks sold
      0.85
      QWK on a public benchmark
      3+2
      Patents & trademarks filed

      A free pilot. Three wins, zero cost.

      One signing meeting, one pilot class, your honest feedback. That's all we ask. We run everything: onboarding, guides and helpdesk. From reply to pilot in under a month.

      Start a free pilot →
      No account needed. Leave your email and we'll send the pilot pack straight to your inbox.