The new head of the Artificial Intelligence Center at the Ministry of Digital Transformation, Roman Kyslyi, is convinced that creating our own domestic LLM model is part of the state’s AI sovereignty.
The new head of the Artificial Intelligence Center at the Ministry of Digital Transformation, Roman Kyslyi, is convinced that creating our own domestic LLM model is part of the state’s AI sovereignty.
He said this in an interview with Channel 24.
«This direction was started by previous leaders when they launched a project with Ukrainian LLM. It is part of what is called AI sovereignty. And sovereignty in the field of AI is, in particular, the development of language,» he noted.
He explained that most of the data on which modern models are trained is English-language content, so «the more we add high-quality Ukrainian data, the more publicly available datasets we create, the better the models understand the Ukrainian language and context.»
In his opinion, the share of Russian propaganda that ends up in training kits is decreasing accordingly simply because there is simply more Russian-language content on the Internet.
Roman Kyslyi believes that this contributes to the development of the Ukrainian language in a technological context, and the results of this activity are already available.
«Previously, Ukrainian was considered a low-resource language — a language with a small number of high-quality datasets. Now, thanks to the efforts of the community, we are moving into the mid-resource category. This is a positive trend,» emphasized the head of the AI Center of Excellence.
He also added that «the data should remain in Ukraine.»
«Personal data, any other. This is sovereign AI. Ukrainian products can be quite competitive for both business and everyday use. Therefore, we are building our own state AI infrastructure AI Factory in a joint initiative with NVIDIA,» said Kyslyi.
According to him, the AI Center of Excellence team is approaching the point where they are not so much interested in the model itself as in the context.
«There are small open models — for example, Gemma 4 from Google — that give good answers out of the box. If these models are connected to the right sources, they will answer with the right context. Therefore, one of my priorities is the development of open data — so that this context becomes more,» Roman Kysly outlined his vision.
As dev.ua previously wrote, in May, the WINWIN AI Center of Excellence at the Ministry of Digital Affairs received a new Chief AI Officer — Roman Kyslyi, an expert with BigTech experience at Deloitte and Visa and a PhD in computer science from KPI.
