Nvidia is discontinuing support for various languages ​​in its NeMo Retriever. From now on, it can retrieve relevant information from datasets at AI prompts, even if that information is available in different languages.

Nvidia is giving its NeMo Retriever a language bath. NeMo Retriever is part of Nvidia’s AI stack and helps companies do this Get Augmented Generation (RAG). At RAG, the request for an LLM is made automatically with relevant background information. For example, if you ask about the status of an order, an LLM won’t just make up the answer. Behind the scenes, the prompt is populated with information from the order database so that the AI ​​model can search for an answer.

Multilingual

RAG is an efficient way to deploy LLMs on current business data without having to train a complex model yourself. NeMo Retriever now ensures that relevant information is added to prompts, regardless of the language in which that information is available. The retriever translates the information into vectors so that the AI ​​can use it efficiently.

Nvidia is convinced that the multilingual nature of the tool is important for large companies. After all, relevant information is not only hidden there in English. The multilingual NeMo Retriever is available as a microservice via the Nvidia API catalog.