Parent Guide · KOAI — Korean AI Olympiad

Before competing on knowledge,
we master the syllabus

KOAI (Korean AI Olympiad) is organized by KITPA and selects Korea's national team for the International Olympiad in AI (IOAI). It is a written competition testing theory and hands-on implementation across machine learning, deep learning, computer vision, and NLP. Unlike a project competition, preparation is about understanding a broad span of AI accurately and solving under time. This page shows that preparation process, in pictures.

Overview Middle · high tracks · written-focused · linked to IOAI team selection Categories, eligibility, dates: check KITPA's notice each year
01 · What's tested

Understanding and implementing four AI domains

KOAI tests not a single project but understanding and implementation across AI, through multiple-choice and short-answer questions. KITPA's official syllabus lists the four domains below, and the scope is updated each year.

Not only theory, but the ability to implement it hands-on with tools like PyTorch. Check that year's official syllabus for the domains and scope that apply per track.

Host: KITPA
Korea IT Promotion Agency
Written-focused
Multiple-choice + short answer
IOAI selection
Feeds the international AI olympiad team

That's why we build the concepts first. Cramming won't get the four domains right. Understand the concept, implement it yourself, then get fluent with the question types — in that order.

02 · Two tracks

Middle and high school differ in format

Even within KOAI, the exam format and stages differ by track. The below reflects the archived 2026 guidelines; categories, eligibility, and dates change each year, so check that year's KITPA notice before applying.

Middle school

One written exam

  • Online written exam, 3 hours
  • 40 multiple-choice + 5 short-answer
  • 1,000 points
  • Focused on syllabus subjects 1–2
High school

Three-stage team selection

  • Stage 1 documents — AI activity portfolio, statement
  • Stage 2 6-hour online exam — 80 MC + 10 short-answer, 1,000 pts
  • Stage 3 national-team interview — English, AI coding, teamwork
  • Top scorers → IOAI international round

The high-school track selects the IOAI national team over three stages, so beyond the exam it also requires documents and an interview. Check which track to enter, and the eligibility and any exemptions, in that year's official guidelines.

03 · Preparation has an order

From concepts to the real thing, in five steps

AI knowledge isn't memorized in one pass. Understand the concept, implement it, get fluent with the question types, then practice under time. The high-school track adds documents and an interview.

Drilling problems without concepts stalls the moment a question varies. Conversely, understanding without implementing and solving by hand means running out of time. Alternating between the two is the knack.

04 · The four domains, what you learn

For each domain, understand and implement

Each domain goes beyond understanding to the level of being able to implement and explain it — because short-answer and interview questions ask "why does this work?"

📐

Foundations · classic ML

The math and data handling AI needs, and the principles of classic ML like regression and classification.

math · scikit-learn
🧠

Neural nets · deep learning

Understand how neural networks learn, and build and train models yourself.

PyTorch
👁️

Computer vision

The principles and limits of image models, implementing classification and detection.

CNN · images
💬

NLP · audio

How models process text and speech, with simple implementations.

NLP · audio

For the scope and the subjects that apply per track, see the competition page and that year's KITPA official syllabus. The syllabus is updated annually.

05 · What a good short answer looks like

Not just the answer — the reasoning and evidence

Multiple-choice tests precise knowledge; short-answer tests the ability to explain why. With the same answer, showing the reasoning is what earns the points.

Answer only — no basis"Accuracy drops." Only the result, with no why and no evidence.
Better — thin reasoning"Accuracy drops because of overfitting." Names the cause, but not how it was verified or how to fix it.
Reasoning and evidence"Training accuracy rises while validation accuracy falls, indicating overfitting, which dropout and data augmentation can ease — the validation curve is the evidence." Observation, cause, evidence, and fix are all there.
06 · The shape of readiness

If only one side is high, it collapses

Check readiness on five axes. Knowing concepts but unable to implement, or fast at solving but unable to explain, stalls on short-answer and interview. Strong prep fills all five evenly.

Concepts Coding Coverage Explain Speed
Solid prep Lopsided prep

The five axes are concepts · coding · solving speed · explanation · syllabus coverage. A diagnostic finds the weak axis and fills it first.

Check (e.g.): if concepts are there but can't be turned into code, add implementation practice; if solving is fast but "why" can't be explained, add short-answer training. Filling the weak axis raises the score most.
07 · So here's the schedule

Concepts first, the real thing later

Count back from the written-exam date in that year's KITPA notice. The red and gold bands below are concepts and implementation, the green is question types and mock exams, and the last is (high-school) documents and interview.

W12345678910111213141516
Understand concepts
AI theory
Implement
PyTorch
Drill question types
MC · short-answer
Mock exams
time management
Docs · interview (HS)
portfolio · English
Concepts Implementation Question types · mock Docs · interview

Fit school exams and assignments first, then set the prep schedule. The concept and implementation bands overlap and repeat with the practice band. Count back the exact written date from that year's KITPA notice.

08 · What your child gains

What stays, apart from the result

This prep is not competition-only memorization. Understanding how AI works and being able to implement and explain it carry straight over to later AI study, projects, and university coursework.

Understanding AI principles

Learning to explain why, not just memorize formulas.

Implementing firsthand

Turning theory into PyTorch code and actually training and validating.

Explaining with evidence

Communicating not just the result but the reasoning, in writing and speech.

Solving under time

Practicing to work a broad syllabus calmly within a set time.

There is little you need to do at home. Listen to which parts of AI spark your child's curiosity, and support steady study time.

Awards are not guaranteed. Categories, eligibility, dates, the tested scope, and IOAI national-team selection differ each year, so always check that year's KITPA official notice (kitpa.org). Korean universities may restrict external-competition records depending on the track, participating/advancing/winning does not guarantee admission, and IOAI national-team selection is a separate track from admissions. For the competition overview see the competition page; for questions, contact your teacher or jc@citcoding.com. Same approach for other competitions: CAC · Technovation · KSEF.

CIT Code Academy · Parent Guide — KOAI preparation KOAI · Korean AI Olympiad (KITPA) · IOAI team selection