System · e-commerce
Automated translation pipeline for 30+ languages
AI Automation PythonCeleryYandex Cloud TranslatePostgreSQLRedis
content change ──► queue (Celery) ──► Yandex Translate ──► 30+ locales ──► store ──► CDN
│
human spot-checks Problem
An e-commerce platform expanding internationally needed its catalog and content localized into 30+ languages; manual translation could not keep pace with content changes.
Approach
Built an automated translation service on Celery task queues integrated with Yandex Cloud Translate — processing content changes asynchronously, storing localized versions for delivery, and keeping humans in the loop for spot-checks.
Result
Continuous localization into 30+ languages without manual steps in the loop, processing content changes as they happened — an early production example of the supervised-automation pattern I now build with LLMs.
Evidence
Described in work history (Amit agency, 2020–2022).
Available for: public discussion
The shape of this system — queue, machine intelligence in the middle, storage, human spot-checks — is exactly the shape of modern LLM automation pipelines. The intelligence got better; the engineering discipline is the same.