Non Profit
A non-profit ran thousands of translation and audio-to-text jobs a year on manual processes, with heavy admin behind every one. We built a workflow factory that removed the manual work and let any team submit a job in seconds.
The challenge
Thousands of translation and audio-to-text submissions ran through the localization team each year, all on manual processes that took significant administrative time. Editorial and marketing needed transcription and translation too, with nowhere self-serve to send it.
What we did
We designed a component-based language factory: a set of microservices tied together by the Blackbird.io workflow orchestrator. One hot folder receives every submission, classifies it by file type and name, and routes it to the right semi-automated workflow. Progress updates itself across Slack and a Trello board as each file moves.
- Translation. Files route to the right translation memory and glossary by domain, through machine-translation and TMS microservices.
- Audio to text. A transcription microservice (Transkriptor and OpenAI Whisper) adds paragraphs, timestamps, and speaker labels.
- File handling. Conversions, word counts, and version archiving run automatically as each job moves.
The result
Manual administrative work dropped by around 40 hours a week. Legacy tools were retired in favor of API-connected equivalents, and people outside localization now submit work themselves. Every step is easy to analyze and adjust, third-party components swap out without a rebuild, and the language assets stay reusable for further workflows and model training.