Hire one more TA and the work shrinks, but the asset you keep is zero.
Materials a TA makes leave with the TA. Materials your system makes stay as each student's weakness map. An ontology that thickens with every exam — that's the asset AXDX builds.
AI cost at cost, zero margin. You pay for my time only.
It's not the students that grew — it's the operations.
If you craft variants yourself, copy grading results by hand, and pick clinic targets from memory — two out of three means it's time. Around 100 students, all three collapse at once.
Material bottleneck
School-specific exam variants are made by hand, every night, every exam season.
Lost answer data
Grading results get copied by hand. Who misses what never becomes data.
Clinic overload
Weakness × timetable × room combinations live in one head. Students slip through.
Feedback delays
Parent reports pile up and get skipped. Re-enrollment wobbles.
Produce, grade, copy it over by hand — that seam is the problem.
Off-the-shelf tools stop at question generation. AXDX removes the re-copying between production and feedback.
- 01
Exam material production
Variant questions + answer keys matched to each school's exam style. Every item carries passage/type/difficulty IDs.
- 02
Per-student tracking
Grading results link to item IDs, so each student's weak types are aggregated automatically.
- 03
Clinic scheduling
Clinic targets and follow-up assignments drafted from weakness data.
- 04
Parent feedback
Per-student report drafts generated automatically. Sent only after your approval, always.
Every exam thickens each student's weakness map.
The product is not the question DB — it's the ontology. Produce items, ingest answer data, grow a per-student knowledge model. An asset that compounds.
Munjaedang — question bank
Every item carries passage, type, difficulty, and testing-point IDs. Starts on a 2,352-item exam DB.
Answer data capture
Exam and homework results link straight to item IDs. The copy-over-by-hand step disappears.
Ontology — the weakness map
Who misses what, when — accumulated. Clinic scheduling, variant recommendations, and parent reports all come off this map.
Solo at 100 students. Crew once you have TAs.
Every engagement starts with a 2-week paid diagnosis (Sprint). Solo plants the workflow inside your own Codex/Claude app — zero extra hardware, right from your phone and PC. Dedicated servers start at Crew.
Paid diagnosis sprint
Pick the one routine that eats the most time. We turn it into working automation in two weeks.
Standing support for one teacher
The full production-to-feedback workflow, planted in your own app and tuned to your exam calendar every month.
Multi-user standing system
Teachers and TAs request via messenger; an always-on server handles it. Built on-site.
Academy-wide build + handoff
Design and build the academy-wide operating system, then hand it over to your staff. (By consultation)
2,352 items, 98 exam runs, zero missing answers.
Not "we'll build it" — it already exists. An asset you use from day one of the contract.
CSAT & mock exam DB
98 exam runs, 2020–2026, reading section (Q18–45). Zero missing answers, full schema validation.
School-exam & clinic DB
Real school-exam and clinic items, structured by passage, type, and testing point.
HS English exam prep + clinics
I produced exam materials and ran clinics for Hanyoung and Gwangmun high schools. I know the labor each school's style costs.
Sample outputs (question packs · wrong-answer clinic sheets · parent reports) shown during consultation.
A TA's salary vanishes monthly. The system accumulates monthly.
This isn't replacing TAs. It's letting you take more students without hiring one more.
- Hire one more TA ₩1.8–2.8M/mo
- AXDX Solo (build + run + AI at cost) ~₩1M/mo
- Headroom you gain +30–50 students
Figures are market estimates. At the first meeting we recompute with your academy's actual TA payroll and tuition. AI subscription stays in your name, billed at cost, zero margin.
More students, collapsing operations?
Tell me in one line which task eats your time right now. We start by checking whether a 2-week Sprint fits.