Illia Polosukhin is a computer scientist and entrepreneur best known as the co-founder of NEAR, a public blockchain platform, and as a co-author of the 2017 paper "Attention Is All You Need," which introduced the Transformer (architecture) architecture that underpins modern large language models. His work spans artificial intelligence, machine learning, and decentralized systems.
Born in Ukraine, Polosukhin studied at the Kharkiv Polytechnic Institute, where he earned a master's degree in computer science. He later worked as a software engineer at several companies before joining Google, where he became an engineering manager at Google Research. During his time at Google, he collaborated with researchers including Jakob Uszkoreit, Lukasz Kaiser, and Llion Jones on the research that led to the Transformer paper, published in 2017. The paper proposed a novel architecture based on multi-head attention and positional encoding, which replaced recurrent and convolutional layers in sequence modeling and became the foundation for subsequent models such as BERT and GPT.
Career at Google and the Transformer
At Google Research, Polosukhin focused on natural language understanding and representation learning. His contributions to the Transformer paper included the design of the encoder-decoder framework and the use of self-attention mechanisms. The paper demonstrated that the Transformer could achieve state-of-the-art results on machine translation tasks while being more parallelizable and requiring less training time than recurrent neural networks. This work has had a lasting impact on the field, enabling the development of generative AI systems and deep learning applications across industries.
Founding NEAR Protocol
In 2017, Polosukhin left Google to co-found Near.ai, an artificial intelligence startup focused on program synthesis, with Alexander Skidanov, a former director of engineering at MemSQL. The project evolved into NEAR Protocol, a blockchain platform designed for high-throughput, user-friendly decentralized applications. NEAR uses a proof-of-stake consensus mechanism and a sharded architecture to achieve scalability. The mainnet launched in April 2020, initially in a restricted proof-of-authority mode, and later transitioned to full proof-of-stake. NEAR raised significant funding, including $12.1 million in July 2019 and $21.6 million in May 2020, with investors such as Metastable Capital and Andreessen Horowitz.
Technical Contributions and Vision
Polosukhin has been a vocal advocate for making blockchain technology accessible to mainstream users. Under his leadership, NEAR introduced features such as human-readable account names, a progressive security model, and a developer-friendly smart contract environment. He has also emphasized the importance of user experience and onboarding, aiming to lower barriers for non-technical users. His background in AI has influenced NEAR's roadmap, including efforts to integrate machine learning tools and decentralized AI applications on the platform.
Recognition and Influence
Polosukhin is widely recognized in both the AI and blockchain communities. The Transformer paper has been cited tens of thousands of times, and its authors are considered pioneers of modern AI. Polosukhin's transition from AI research to blockchain entrepreneurship highlights the convergence of these fields, as decentralized networks increasingly rely on AI for automation and data processing. He has spoken at major conferences and contributed to discussions on the future of web3 and decentralized governance.
Controversies and Challenges
NEAR has faced criticism and security incidents. In 2022, Skyward Finance, a launchpad built on NEAR, was exploited for approximately $3 million due to a flaw in its treasury redemption contract. In 2023, a report by Messari highlighted concerns about validator decentralization, noting that about 35% of staked tokens were hosted on Amazon Web Services, leading to a low operational Nakamoto coefficient. In 2024, NEAR's official X account posted anti-crypto messages that were later revealed to be a marketing stunt, drawing backlash from the community. These events have tested Polosukhin's leadership and the project's resilience.
Legacy and Future Directions
Polosukhin continues to guide NEAR's development, focusing on scalability, interoperability, and user adoption. His dual expertise in AI and blockchain positions him as a key figure in the emerging field of decentralized AI. As of 2025, NEAR remains one of the prominent layer-one protocols, with ongoing efforts to improve its infrastructure and expand its ecosystem. Polosukhin's work exemplifies the cross-pollination of ideas between AI and blockchain, and his contributions are likely to influence both industries for years to come.