ZHADIGERAI: AI SERVICES FOR KAZAKH LANGUAGE

Abstract

This project addresses the underrepresentation of the Kazakh language in modern AI services by designing and deploying a comprehensive web platform offering 11 AI services tailored for the Kazakh language, including ext-to-Speech, Speech-to-Text, Question Answering, Automatic Summarization, Machine Translation, Named Entity Recognition, Optical Character Recognition, Image Captioning, Large Language Model, KazCLIP, and VideoCLIP. The system features a React frontend and Spring Boot backend with secure user authentication and efficient resource management. For models serving we utilized Triton Inference Server framework. The platform follows the full cycle of designing, implementing, and evaluating a computing-based solution.