NetKoçum

Overview

netkocum.com is a platform where a student preparing for Türkiye's university entrance exam (YKS) logs what they study, and the system works out which topics are weak and what to focus on this week. Mock exams, a mistake box, a review schedule, a weekly plan and a bank of 7,600 original questions live in one place, and the student's coach sees the same screens. I built every layer myself, from idea to launch.

• Rules engine: 12 risk rules run over a snapshot of the student (topic states, the last 35 days of logs, mock exams, mistakes) and raise alerts such as accuracy bands, overdue reviews, blank rate and net drops, each with a suggested next step.
• Automatic weekly plan: picks overdue reviews, low-accuracy topics and the next new topics by their weight in the exam and spreads them over the week; the coach edits and publishes the draft. A catch-up estimate compares the last four weeks' pace with the pace needed before exam day.
• Official resource catalogue: 177 free MEB and ÖSYM publications with 2,089 topic mappings, each link opening the PDF at the page where the topic starts. No content is copied; only the link and page number are stored.
• Question bank: 7,600 original questions with explanations across 25 TYT and AYT subjects. Topic tests favour unseen and previously missed questions; there are grade-level (9-12) tests, a daily challenge and a speed round.
• Two curricula side by side: the 2018 programme (YKS 2027) and the new Maarif model (YKS 2028+). MEB's 2026 programmes were read subject by subject and each of the 487 topics was mapped; a student's programme follows their exam year.
• KVKK (Turkish GDPR): privacy notice and explicit consent, parental consent under 18, data export and account deletion; coach access is visible to the student and logged. Data stays on a server in Türkiye.

Screens

01

Problem

02

Approach

03

Architecture

04

Key decisions

1. 01

   Analysis in a pure rules engine

   If the rules touched the database or the clock, every test would need a data setup. Because the engine is pure, a rule's behaviour is tested with a single snapshot, and the plan suggestions come out of the same engine.

2. 02

   Links and pages, not content

   Copying official publications raises both copyright and staleness problems. Only the link, the topic and the physical page number are stored; the PDF opens on its own site at the right page, and broken links are checked in bulk from the admin panel.

3. 03

   Topic IDs shared across curricula

   Building the new curriculum as a separate topic tree would have split the question bank, resources and student progress in two. With shared IDs, a programme differs only in its topic list and grades, and narrowed topics carry a scope note.

4. 04

   Refuse to boot on inconsistent data

   The curriculum, question bank and resource catalogue are JSON files that ship with the code. Each is validated at startup; an unknown topic, a wrong grade or an out-of-range page stops the app from booting. The error surfaces in the build, not in front of a student.

Outcome

What I learned

“In an education product the hardest part isn't the code, it's the correctness of the data. One wrong page number or answer key and the student stops trusting it. So every dataset is tied to a verification script and a startup check, and when the curriculum changed I read the official documents first and changed the code second.”