Ukraine and Kyivstar Collaborate with Google to Create National Ukrainian Large Language Model

The Ministry of Digital Transformation of Ukraine and Kyivstar, the nation’s largest mobile operator under telecoms group VEON, have partnered with Google to develop a groundbreaking national large language model (LLM). This innovative initiative marks a significant step forward in integrating cutting-edge artificial intelligence tailored specifically for the Ukrainian language, its culture, and historical context. Utilizing Google’s Gemma and Google Cloud’s Vertex AI, the project aims to establish a robust technical backbone for AI developments across critical sectors, including healthcare, education, and finance.

Revolutionizing AI with Ukrainian Context

The Ukrainian LLM, built on Google’s Gemma model, will uniquely address linguistic, cultural, and historical nuances of Ukraine that have often been sidelined in global AI models. Kyivstar and the Ministry of Digital Transformation emphasize the importance of training the LLM on curated, Ukrainian-specific datasets while retaining sensitive data within the country. This approach not only ensures data sovereignty but also taps into the unique attributes of the Ukrainian language and culture. According to Danylo Tsvok, Chief AI Officer of the Ministry, the emphasis will be on linguistic precision and ethical considerations to minimize risk.

Google Cloud’s Vertex AI infrastructure provides the project with the computing power required for large-scale training and refinement. Commenting on this partnership, Krzysztof Kaziów, Director of Customer Engineering for Central and Eastern Europe at Google Cloud, emphasized the importance of localized AI in enhancing the digital experience for Ukrainian users. This collaborative endeavor is expected to pave the way for more sustainable and inclusive AI developments in Ukraine.

The Role of Sovereign AI in Nation-building

A key aspect of this initiative is its cultural and national significance. Historically, many Ukrainian datasets have been influenced by foreign narratives, particularly Russian, leading to biases in existing language models. The creation of a sovereign LLM aims to rectify this imbalance, providing a model that accurately represents Ukraine’s linguistic and historical heritage. VEON CEO Kaan Terzioglu highlighted that this localized AI will empower Ukrainian businesses, institutions, and individuals to embrace augmented intelligence that genuinely understands and speaks Ukrainian.

This development also aligns with VEON’s broader strategy of promoting local-language AI tools in its markets, exemplified by prior projects like KazLLM in Kazakhstan and an Urdu LLM in Pakistan. For Ukraine, this initiative transcends technological innovation by contributing directly to its national identity and independence in the digital age.

Applications and Future Potential

The Ukrainian LLM is expected to serve diverse applications, including regulatory analysis, educational tools, healthcare advancements, and financial services. Developers will optimize Gemma for Ukrainian by refining its tokenizer and creating benchmarks tailored to domain-specific use cases. The long-term goal is to enable flexible and intuitive AI solutions that meet the country’s unique needs.

As Ukraine takes steps to reinforce its digital and linguistic sovereignty, this AI project not only showcases the power of collaboration but also sets a benchmark for leveraging technology to enhance cultural identity. By addressing historical biases and prioritizing local contexts, Ukraine is charting a path toward an AI-driven future that remains deeply rooted in its heritage.