NMT.academy

The idea came from our history teacher. He asked us to build a small platform to help his students prepare for the NMT. For about a year it was a simple LMS.

Then we decided we wanted to earn more from it and rebuilt it as an adaptive platform, where everything adapts to the student, because we live in the era of customization.

Not because I didn’t prepare. I get so afraid that it almost turns into a panic attack. My nose bleeds when I write exams, and sometimes I can’t even see the sheet in front of me. I saw a psychologist about it as a teenager.

That forced me to work out early that an exam doesn’t only measure what you know. It measures how you perform inside a narrow format, under time, once. That is a separate skill, and almost nobody teaches it.

This is what pushed me to build NMT.academy: so that students walk into the exam knowing the subject and knowing how to handle the exam itself.

A topic does not count as closed until the student has seen 85% of its question bank and answers 90% of it correctly — and that is checked on fresh attempts, not on the questions they just did.

Questions are tagged by what they actually test: 6,688 of the 7,407 history tasks carry those tags today. A student’s answers add up to a map of their gaps, and the lessons, notes and tests follow that map and the score they’re aiming for. An exam simulation estimates where they are heading.

Explanations are written by us, not generated: 4,018 of the tasks carry one today. AI does the authoring work behind the scenes — tagging questions and drafting wrong options — not the teaching. The tagging, the answer base and the algorithm are ours, and the engine isn’t tied to a subject or a language. A second course is already built, and an English version works.

Teachers get a cabinet that shows each student’s gaps, not the class average.

We took the NMT ourselves. Ihor got two 200s.

Maksym, who now runs our social media and course production, had a tutor for all four of his NMT subjects in 2026. Only in history did our platform come on top of the tutor.

The NMT is scored from 100 to 200. He scored 194 in history, 52 points above the national average. With a tutor alone he got 172 in Ukrainian, 173 in English and 140 in maths, after about 340 hours of maths lessons: 8 to 26 points above the average. The only difference was the platform.

20 of the 22 students who aimed for 190+ out of 200 got it, and the median in that group was 194. In history that is 52 of 54 test points, or 96% of the exam right. The national average in history in 2025 was 141.6, about 40% of the exam.

It’s a small group, and we say so. The big exam-prep companies report 183 on 60,000 students and 188 on 4,000. Our claim is narrower: did the student reach the score they aimed for.

This year’s cohort is 49 students so far. 48 of them set a target: the median is 188, 41 aim for 180 or more, and 16 want a perfect 200.

A group of 22 top scorers invites the obvious objection, so this year we measure the whole class instead. Across 49 students and 114,000 answers, the median one now makes 43% fewer mistakes on the same material, and two in three mistakes are fixed on the next try.

I came up with all of it: the name, the green, the mascot. The mascot was meant to be the logo and turned into a mascot. Green because it fits and is calm enough to study with.

I also manage our B2B leads and our first clients. My co-founder Ihor builds the platform and the algorithm.

We decided not to go to the B2C market. We bring in teachers, and they stay with us for a long time, because we raise their students’ results a lot. That’s where our first paying students came from.

Teachers have signed on and are starting to build their own courses on our platform. Three tutors want to join. Exam-prep centres are next.

The biggest struggle is convincing a lot of people that an algorithm can really prepare a student better than a person, or at least help that person. Teachers have to see that we’re strong enough to compete with, and that we make their work easier.

On top of that, we’re showing the market something new. Nobody has done this before, at least in Ukraine. Inside the team, the hard part was our arguments.

I’d build the adaptive platform from day one, and go B2B from day one. People need a lot of customization to get high results today.

Next: foreign markets, like the SAT, or getting inside a large number of private schools.