W17 – Pipeline: pre-edit, human review, AI review, back-translation as optional stages
The localization pipeline wraps the core translate step with optional stages, each independently toggleable per engine and overridable per job:
Pre-localization AI edit. An AI agent cleans the source payload before translation, so a single source error doesn't propagate across every target locale.
Post-localization human edit. Sends the translation to a qualified human translator. The job pauses on a webhook until the edit returns.
Post-localization AI review. Reconciles the human output against the engine's glossary, brand voice, and instructions.
Back-translation check. Translates the final output back into the source locale and compares. The agent flags semantic drift by severity and auto-adjusts on major or critical drift.
Turn on what you need, leave the rest off.
Full changelog: lingo.dev/changelog
Replies
I've worked with similar processes, and I think the pre-edit stage often gets overlooked. In my experience, cleaning up the source content early can reduce review time later and improve the final translation quality.
I appreciate the layered review approach. For me, combining human judgment with AI validation creates a stronger quality assurance process than relying on either one alone.