All projects

LLM translation cleanup

Cleaning up product translations inside Windly

AbearApr 2024 – Jun 20243 mos

Context

Windly surfaces overseas products to domestic buyers, so translation quality is product-page quality.

Problem

Option names, product titles, and tags were translated inconsistently, which degraded the buying experience.

How I solved it

  1. Switched the approach from rule-based to LLM-based cleanup, designing and building the pipeline on Langchain
  2. Proved it internally with a Streamlit MVP first, then built the serving backend

Result

  • The cleanup pipeline shipped to production inside Windly

Impact

Product copy now reaches the storefront cleaned to a consistent standard.

Tech

PythonLangchainStreamlitFastAPI
NextCategory recommender v2New categories without retraining