Сервіс генерації синтетичних e-commerce датасетів на основі LLM для практичного навчання дата інженерів

Abstract

This work is devoted to the development of a platform for generating synthetic e-commerce datasets based on large language model simulations for practical training of data engineers. The platform provides access to generated data through a REST API and a webhook-based push system, enabling integration into existing training pipelines. A comparative analysis of existing dataset generation approaches was carried out. It has been found that the proposed LLM-based approach produces statistically realistic data with configurable complexity, making it suitable for practical data engineering training without violating data privacy requirements.