ASTANA – The Institute of Smart Systems and Artificial Intelligence (NU ISSAI) at Nazarbayev University has launched Oylan, a multimodal AI model capable of processing both images and text in Kazakh, English, and Russian, reported the NU ISSAI’s press service.
Photo credit: oylan.nu.edu.kz
To train Oylan, NU ISSAI researchers created the largest AI dataset in Kazakhstan, comprising over 10 million images and 50 million question-answer pairs. The dataset is tailored to the country’s linguistic, cultural, and economic needs, enabling Oylan to be applied in education, healthcare, industry, and finance.
“The dataset covers a wide range of domains such as image captions, visual questions and answers, optical character recognition, document analysis, understanding charts, graphs and tables, problem solving in various fields such as math, geometry, physics and more,” noted Askat Kuzdeuov, the lead data scientist.
Despite limited computational resources, the team developed efficient training techniques to optimize Oylan’s performance. The model is now available for public testing on the ISSAI Playground, featuring voice interaction for a more intuitive experience. Developers can also access Oylan’s API to integrate it into various applications.
