The portal had to support around eight languages so that editors could maintain a single original version of each article while localizations are generated and updated without manual duplication.

News, events, and jobs share topics and sections but have different data structures and filter logic. The data model had to let content intersect through shared categories without forcing separate structures.

Ad placements were managed without any centralized system. The client needed a tool where ad zones, page- and category-level targeting, and creatives could all be managed without touching code.

News editors, event managers, job managers, advertising admins, and external partners all needed clearly different access rights.

The system had to integrate with DeepL while still letting editors decide what to translate automatically, where to edit by hand, and how to track localization freshness.

Some external companies needed to publish jobs themselves — through the Back Office or the API. That required dedicated roles, quotas, and moderation hooks.

We started by defining all content types, then for each type defining required fields, filters, and shared categories — so a user could browse, for example, events and news on the same topic without duplicating the underlying structure.

We set up entities for News, Event, Job, Category, Location, and Employer in Directus. Categories and tags are used simultaneously by different content types, so the team could manage the topic tree in one place and immediately affect how the entire content stream renders.

We implemented multilingual handling through a service that sends base-language text to DeepL, receives the translation, and fills the corresponding fields for the other languages. Editors can flag records as "needs review," and the system stores when each translation was made and from which source version.

We laid out the Back Office as clear sections: news, events, jobs, advertising, users, and partners. Each module has its own role setup — news editors don’t see advertising settings, job managers work only within their own section, and administrators control the category structure and the language list.

In the backend, we designed a model where an ad zone is bound to a page type, category, language, and on-page position. That lets advertising managers turn specific blocks on and off, swap creatives, and set display priorities — without touching frontend code.

For external companies, we created separate accounts with restricted permissions: creating and editing their own job postings, viewing applications, and integrating through the API. At the API layer, we implemented access keys, quotas, and basic rate limits — to keep the platform stable and avoid overloading moderation.

In Directus, we enabled user-action auditing and added custom logs for publication events, localization updates, and ad-zone configuration changes. Logging also helps track failures from DeepL calls and partner API requests, so the team can react quickly when something breaks.

from a single source — auto-localized, with status indicators for readiness control

to publish content, including all localized versions

for banner management — ad zones, targeting, priorities, and creatives are all managed in the Back Office

to launch a new ad campaign, because the manager assembles it from pre-built zones and display rules

for partners — Back Office or API — letting external companies post jobs on their own

of Back Office actions are audited; every action and every change is logged