AI-Assisted Localization Pipeline for the Van-Rental Site
A van/mobility rental brand
Overview
The internationalization companion to the heggo-clone van-rental site: a Python pipeline that translates the site’s UI/content strings into Romanian (and beyond) using an LLM, processing a localization CSV into a translated output CSV.
Why It Exists
The marketing site needed accurate, consistent multi-language copy without hand-translating every string. An automated, resumable pipeline keeps translations consistent and lets the work be re-run as content evolves.
What We Built
A translate_localization.py script that reads Localization Heggo.csv, translates each entry via the OpenAI API, and writes Localization Heggo_RO.csv. It runs translations concurrently with asyncio/aiohttp (bounded to a max-concurrency pool) and persists progress to a JSON checkpoint file so an interrupted run resumes from the last completed position rather than restarting or re-billing completed strings.
Technologies & Approach
Python with asyncio + aiohttp for bounded concurrent throughput, the OpenAI client for translation, and a checkpoint/resume design for resilience and cost control. Dependencies are minimal (openai, python-dotenv, aiohttp).
Outcome / Impact
Produced a reliable, resumable localization workflow that turns a source string set into translated locale files automatically, reusable for any CSV-driven i18n need, and directly powering the multilingual content for the paired marketing site.
Capabilities Demonstrated
- LLM-driven content localization at scale
- Resumable, checkpointed batch pipelines
- Bounded-concurrency async API processing in Python
- Cost-aware automation of repetitive translation work