KEP.uz
Programming practice, competitions and automated judging.
Overview
A programming platform with algorithmic problems, practical software projects, contests and duels. React workspaces connect to Django services and an online judge that executes submitted code in isolated environments.
The challenge
The platform supports several task formats and scoring rules. Code execution requires isolation, while the workspace needs consistent feedback on test results and scores.
Engineering decisions
One modular judging pipeline
Problem packages use checkers, validators, generators and reference solutions. The engine handles standard tasks, partial-score subtasks, interactive protocols, C++ graders and function exercises, while returning useful test and score summaries.
Test working software
Practical projects extend the platform into frontend, backend, Python and DevOps. Submitted interfaces are exercised in a real browser with Playwright; backend projects are checked through HTTP contracts with isolated test state.
Connect practice to competition
Timed contests, virtual participation, head-to-head duels, standings and ratings form a continuous learning loop. React workspaces bring the statement, editor, custom runs and submission feedback together.
Keep growth manageable
Query-budget regression tests protect catalogue and standings performance. Background judging starts after the submission transaction commits, and the newer judge can coexist with legacy problem packages during migration.
What it delivers
Public statistics recorded on 26 October 2026: 9,564 users, 2,101 visible problems, 374 contests and 436,057 submissions. Keppy World adds a multiplayer learning environment with server-verified puzzles and persistent progress.
Sources & context
Platform totals come from the public KEP.uz statistics API, observed 26 October 2026 and cached by the source for 24 hours. Features are supported by source review; founder status is also corroborated by IT Park.
KEP.uzPublic platform snapshot · 26 Oct 2026