This article presents the development of an intelligent system for searching and recommending organizations in the service sector. The software allows users to enter queries in natural language, after which the query is classified and the most relevant results are issued from the Yandex Geo-Reviews dataset. The software architecture is described, including a frontend on the React framework, a backend in the Golang language using FastHTTP, and an ML service on the FastAPI platform. SBERT models are used for text analysis and the CatBoost classifier is used to form a set of preference classes for organizations. The article also discusses methods for interaction between system components at the conceptual level and at the level of data exchange protocols, which allows other developers to use the proposed methodology to create similar software solutions.

machine learning, recommender systems, software development frameworks, data exchange protocols

Mathematical modelling in actual problems of science and technics