We have built a proprietary retrieval and embedding model, Oasis. Oasis creates an automatic taxonomy from your dataset and uses this taxonomy to create interpretable vector embeddings. In doing so, you have a RAG system that does not need a lot of prompting and almost no fine-tuning, that is fully interpretable, and leaves an audit trail.
We believe this solves a lot of issues with getting RAG beyond 80% demos and into full enterprise applications, but we’re curious if you see this too. Kaspar and Bekk will show you the playground and our API!
