Yandex: Neuroexpert

2025: Product Announcement

Yandex launched in April 2025 a beta version of the Neuroexpert intelligent service, designed for automated work with corporate data. The new product based on artificial intelligence is capable of processing text documents, presentations, tables, as well as audio and video materials, forming a structured knowledge base on their basis. The tool allows users to quickly find relevant information by asking questions about downloaded data in natural language.

According to Yandex, users can upload links from the Internet and files of various formats to the service, and then ask questions in dialogue mode. "Neuroexpert" analyzes content and generates responses based solely on downloaded sources without adding extraneous information.

In the beta version of the service, users can upload up to 25 files and links to one knowledge base. Once the beta is complete, this restriction will be removed. The created knowledge bases can be shared with other users via a link, which resembles the functionality of cloud storage.

In addition to searching for information, Neuroexpert can retell the content of documents, draw general conclusions or comparisons, and also create texts in the desired style based on the downloaded data. The service is available to all users, including Yandex 360 customers.

Neuroexpert uses several company technologies to extract information from various types of documents. The visual-language model finds the necessary data in graphs and diagrams, speech recognition technology is used to process audio and video, and an optical character recognition system works with text on images.

Answers to user questions are generated by the YandexGPT 5 Pro language model - the latest version of the company's large language model. The integration of neural network knowledge and information from downloaded files is implemented using the RAG system (Retrieval Augmented Generation).