Ragout

Leading LLMs are powerful platforms for language learning.

But they still aren’t on-par with an elite language teacher. They lack the nuance, sensitivity, and hard-won pedagogical strategies that characterize high quality language instruction.

Ragout is a Python pipeline that combines the flexible content generation capabilites of LLMs with proprietary content of a team of veteran French teachers.

Given a brief lesson description, the tool searches over a corpus of hundreds of bilingual unstructured teacher notes, video transcripts, and exercise materials.

It then ranks and filters retrieved documents based on relevance, and passes them to a chain of LLM prompts designed to generate intermediate artifacts such as scenarios, examples, and subskills.

The final prompt generates a narrative lesson script, which is written into Google Docs along with the upstream intermediate artifacts.

Ragout helped the 6pm team bootstrap an ambitious content pipeline, significantly reducing the time required to develop and edit over 75 lesson scripts.