Kazakhstan Develops Homegrown AI Model for Turkic Language Speech Recognition

Kazakh startup Cybernet AI has unveiled Central Asia’s first automatic speech recognition (ASR) model designed specifically for Turkic languages, marking a major milestone in regional artificial intelligence development.

According to the developers, the technology is a fully original product rather than an adaptation of Western solutions. The project was implemented with strategic support from the Ministry of Digital Development and Artificial Intelligence and Astana Hub, while Microsoft’s GPU infrastructure was used to train the model.

Astana Hub Managing Director Tanat Uskembayev said the initiative reflects Kazakhstan’s shift from being a consumer of global AI solutions to becoming a developer of internationally competitive technologies.

The Cybernet AI system can accurately recognize not only formal speech but also mixed-language patterns commonly used in everyday communication. It supports Kazakh, Turkish, Uzbek, Kyrgyz, Azerbaijani, and Tatar, as well as Turkic-Russian code-switching.

Unlike many international ASR models focused primarily on English, the new solution effectively captures Turkic phonetics, accents, and language transitions. The model is designed for use in banking, government services, telecommunications, and contact centers, offering high accuracy even in noisy environments.