COU
Summary
An open-source project built so that anyone can deploy and run an online judge in their own environment. Solutions submitted in the browser are judged by an isolated judging server, and practice problems, contests, and a community come together in one service. It grew from a 3-person project into a 10-person startup.
- Role
- FounderPM
- Outcome
- 3→10-person team2022 SW Talent Festival award
From a project to a startup
COU, short for "Coding Online-judge platform Univ.", is an open-source online judge platform. Development began in December 2021. The project started with the goal of building an online judge that anyone could customize for their own needs and deploy and operate in their own environment.
Centrally operated online judges such as Baekjoon could not be modified to suit individual needs or deployed and run in a separate environment. We set out to build a coding assessment platform that users could extend with the features they needed and operate in their own environment. Operators can set the service name and decide whether to allow registration, as well as configure problems, categories, and contests. For example, they can group selected problems into a category or schedule a contest for a time of their choosing. With small teams in mind, we considered both how to make features easy to modify and how to keep installation and deployment straightforward.
I began by writing a proposal and recruiting two teammates, and the three of us started development together. As development progressed, I grew the team to ten people. Once the platform was reasonably complete, I registered the business and turned the project into a startup. We also published a related paper. The platform was also selected as an outstanding work at the SW Talent Festival.
The solving experience at the core
Solving a problem involves repeatedly reading its statement and sample input and output, writing code, and submitting it to check the result. To support that flow, we placed the problem statement and code editor side by side on a single browser screen. Users can refer to the examples on the left while writing code on the right and submit it from the same screen. The editor supports multiple languages, syntax highlighting, and theme switching. Submitted code runs on a separate judge server to determine whether it passes, and users can check the result in the judging status view.
A solution is not always finished in a single submission, so users also needed a way to retain code they were still working on. The code for the same problem can differ by language, so we added a draft module that automatically saves code separately by user, problem, and language. When users resume solving, the module loads the draft matching those conditions and updates it only when the ID and checksum in the save request match that draft. For coding tests, we also supported private submissions so that participants could not view one another’s submitted code.
Problems, practice, and personalization
The problem list helps users choose problems to practice and see how much of a set they have completed. Operators can group problems into categories, and users can check their progress through the completion rate for each category. Problems have tags and difficulty levels, and users can find them through search, difficulty filters, or random selection. Within the same service, users can view contests (coding tests) and submission status, write questions and posts in the community, and receive notifications.
Stack and deployment
Because the screens and features needed to be adaptable to different operating needs, we separated the frontend and backend through an API. The Vue screens communicate with Django APIs, allowing the UI and server functionality to be modified separately.
The backend is organized into modules for accounts, problems, submissions and judging, contests, drafts, email verification, announcements, and the community. On the frontend, CodeMirror provides code editing, KaTeX and ECharts display formulas and charts, and i18n supports multiple languages. PostgreSQL and Redis handle data storage and processing.
To deploy and operate the service in their own environment, users needed the data storage and judging environment alongside the web application. We packaged the web application, PostgreSQL, Redis, and judge server as separate containers and grouped them with Docker Compose so they could be started together with one command. Submitted code runs in a separate judging container, with restricted privileges to reduce the risks of executing malicious code. In the GCP deployment, we confirmed the working flow from problem creation to code submission and viewing judging results.
Role
As PM and founder, I led product direction, team operations, and external presentations.
During development, we also addressed problems that appeared in the screens after features were added. My work included fixing uploaded images overflowing their display area and comment-editing buttons extending beyond their container. I also worked on details of the usage flow, such as requiring verification again after an email address was changed.
Presenting as school representative at the Gangwon SW-university workshop
In December 2022, I presented COU as my school’s representative at the Gangwon-region SW-oriented University cooperation workshop during the Korea Software Congress. I introduced COU as a coding assessment platform that users could customize for their needs and easily deploy and operate in their own environment.
Selected as an outstanding work at the 2022 SW Talent Festival
I participated as my school’s representative in the 2022 SW Talent Festival, hosted by the Ministry of Science and ICT and organized by IITP and the Association of Software-Centered Universities. COU was selected as an outstanding work at the event, which brought together SW-oriented universities from across the country. I ran a booth, demonstrated the project, and presented it.