Illia Polosukhin

Illia Polosukhin was born in Ukraine and grew up in Kharkiv, the country's second largest city and a longstanding center of engineering education. Accounts of his early years describe a child drawn to computers well before secondary school, and he has said in interviews that he began programming at a young age, an origin story common among the engineers who later built the infrastructure of modern machine learning ^[4].

He pursued formal training in computer science in his home city, studying applied mathematics and computing at Kharkiv Polytechnic Institute, one of Ukraine's oldest technical universities ^[4]. The rigorous mathematical grounding typical of post-Soviet engineering curricula shaped his approach to software: he gravitated toward problems where theory and large-scale systems engineering meet, a combination that would later define both his research at Google and his blockchain work.

For readers searching for basic Illia Polosukhin facts, the essentials are quickly stated. He is a living Ukrainian-born engineer, catalogued on Wikidata under the identifier Q110864222, whose public record begins in earnest with his industry career in the 2010s rather than with a conventional academic trajectory of doctoral study and professorships ^[1].

Polosukhin's route to prominence ran through Google Research, which he joined in the first half of the 2010s after earlier industry work in machine learning. At Google he worked on natural language understanding, contributing to systems that helped the search engine answer questions directly rather than simply returning links ^[4]. This was a period when deep learning was rapidly displacing older statistical methods across the company, and question answering sat at the frontier of that shift.

He also contributed to Google's open source machine learning ecosystem during the era when TensorFlow, the company's deep learning framework, was becoming a standard research tool. Colleagues from that period describe him as a fast-moving engineer with an appetite for ambitious projects, someone who preferred building working systems to publishing incremental papers ^[4].

The question of who was Illia Polosukhin before the Transformer paper has a straightforward answer: a working research engineer, one of thousands inside Google, distinguished mainly by his focus on making neural networks read and answer questions about text. That specific problem, and its stubborn computational bottlenecks, set the stage for the collaboration that made his name.

In 2017 Polosukhin was one of eight Google researchers credited on "Attention Is All You Need", the paper that introduced the Transformer architecture ^[1]. The work, presented at the Neural Information Processing Systems conference, proposed dispensing with the recurrent and convolutional networks that then dominated sequence modeling and relying instead entirely on a mechanism called self-attention. The change allowed models to be trained in parallel across enormous datasets, removing a bottleneck that had constrained the field for years ^[1].

Journalistic reconstructions of the project describe an unusually flat collaboration in which the authors were listed with a note stating that contribution order was random, an arrangement reflecting how thoroughly the ideas had been developed collectively ^[2]. Polosukhin, by these accounts, was among the early participants in the discussions that shaped the architecture, having been motivated by the practical need to make question answering fast enough for real products, and he left Google while the work was still being completed, remaining a credited author ^[2].

The paper's influence exceeded anything its authors anticipated. The Transformer became the foundation of the large language models that later transformed the technology industry, including OpenAI's GPT series, and by the mid-2020s it ranked among the most cited computer science papers ever written ^[3]. In 2024 the surviving group of co-authors, Polosukhin among them, appeared together on stage at NVIDIA's GTC conference in a panel moderated by chief executive Jensen Huang, a public acknowledgment of how central their 2017 work had become ^[5]. Any serious Illia Polosukhin biography treats this paper as the hinge of his career, even though he spent barely a decade in pure AI research before changing course.

Polosukhin left Google in 2017 and, together with fellow engineer Alexander Skidanov, a former ICPC programming competition medalist who had worked at Microsoft and the database company MemSQL, founded a startup called NEAR.ai ^[6]. The company's original goal was teaching machines to write code from human descriptions, a research program known as program synthesis. The founders soon ran into a practical obstacle: paying contributors around the world for training data was slow and expensive through conventional financial systems.

That frustration pushed the team toward blockchain technology, and in 2018 the company pivoted to building NEAR Protocol, a proof-of-stake blockchain designed to scale through a technique called sharding, which splits the network's workload across parallel segments ^[6]. The project raised substantial venture funding during the boom in blockchain investment, and its mainnet went live in 2020, with the network transitioning to community governance later that year ^[6].

Among Illia Polosukhin achievements, NEAR is the one with the most direct commercial footprint. The protocol became one of the more widely used smart contract platforms of the early 2020s, marketing itself on usability features such as human-readable account names and low transaction fees. In 2022 the NEAR ecosystem attracted one of the larger funding rounds in the sector, led by prominent venture firms, at a moment when the founders were positioning the chain as an entry point for mainstream developers ^[6]. Polosukhin served as the project's most visible public spokesman, a role he retained as the NEAR Foundation and a distributed set of core teams took over day-to-day development.

The explosion of interest in generative AI after 2022 pulled Polosukhin back toward his original field, this time with a distinctive argument: that artificial intelligence should not be controlled by a handful of large corporations. He began advocating what he calls user-owned AI, meaning models and assistants whose data, incentives, and governance belong to their users rather than to centralized platforms, with blockchain infrastructure providing the accountability layer ^[3].

In 2024 the NEAR Foundation announced renewed investment in AI research under the NEAR.AI banner, effectively reviving the company's founding mission with Polosukhin at the front of the effort ^[3]. He became a frequent commentator on the economics of the AI industry, arguing in interviews and conference appearances that open source models and decentralized ownership are necessary counterweights to the concentration of computing power ^[5].

This phase of his career has made him an unusual figure: a credited co-inventor of the technology behind the AI boom who spends much of his time warning about how that boom is being governed. Whether user-owned AI becomes a durable movement or remains a niche position, his advocacy has given the debate a spokesman with rare technical standing ^[3].

Polosukhin keeps his private life largely out of public view, and responsible accounts of living persons respect that boundary. What is well documented is his continuing identification with Ukraine. After Russia's full-scale invasion in 2022 he spoke publicly about supporting Ukrainian engineers, and the NEAR ecosystem was among the technology communities that organized aid and employment channels for developers affected by the war ^[6].

He has been based in the United States for much of his career, having relocated there for his work at Google in the early 2010s, and he maintains an active public presence through conference talks, podcasts, and social media, where he posts about AI research, blockchain engineering, and open source software ^[4]. Unlike many technology founders of comparable prominence, he has retained a reputation as a hands-on engineer, and he continues to discuss technical detail in public forums rather than delegating it entirely to staff.

It is unusual to weigh the legacy of someone still in mid-career, but the outline is already clear. The Transformer architecture that Polosukhin helped create in 2017 reorganized the entire field of artificial intelligence, and its citation record and commercial descendants guarantee the paper a permanent place in the history of computing ^[1]. Every major language model deployed in the mid-2020s descends from that design, a fact repeatedly noted in retrospective coverage of the paper's authors ^[2].

His second act with NEAR Protocol gives him a distinct legacy in decentralized computing, where the project's sharding design and developer-friendly account model influenced how later blockchains approached scalability ^[6]. The eventual verdict on NEAR depends on the long-term fate of the blockchain industry itself, but Polosukhin's willingness to leave one of the most celebrated research groups in AI to build financial infrastructure marked him as a founder guided by his own priorities rather than by prestige.

For anyone assembling an Illia Polosukhin biography decades from now, the through line will likely be ownership: of ideas, in a paper whose authors deliberately randomized their credit order, and of technology, in his sustained argument that both money and intelligence should belong to the people who use them ^[3]. Few engineers of his generation have left fingerprints on two separate foundational technologies. He is one of them.