If you’ve used ChatGPT, Google Gemini, or basically any modern language model in the last few years, you’ve already been touched by the work of Illia Polosukhin.
I’ve spent a good amount of time going down this rabbit hole, and honestly, his story is one of the more fascinating ones in tech right now. We’re talking about someone who helped reshape AI at its core and then walked away to build blockchain infrastructure. This piece covers who he is, why that pivot matters, and what he’s actually building.
Illia Polosukhin at a Glance
- Full Name: Illia Polosukhin
- Nationality: Ukrainian
- Known For: Co-authoring Attention Is All You Need
- Former Employer: Google
- Current Role: Co-founder of NEAR Protocol
- Industry: AI + Blockchain
Key Takeaways
- Illia Polosukhin co-authored the Transformer architecture that powers every major AI model in use today.
- He didn’t abandon AI for blockchain. He built the NEAR Protocol because no existing infrastructure could solve a real problem he was already facing.
- NEAR’s focus isn’t just another L1 play. It’s specifically designed as Web3 infrastructure for the agentic economy.
- His vision for decentralized AI isn’t ideological. It’s an engineering argument: open source AI infrastructure is more trustworthy and more regulatable than centralized black boxes.
- With $540M+ raised, 100M+ accounts, and Confidential Intents live on mainnet as of March 2026, the project has moved well past the whitepaper stage.
Who is Illia Polosukhin?
Illia Polosukhin is a Ukrainian computer scientist, AI researcher, and entrepreneur best known for co-authoring Attention Is All You Need and co-founding NEAR Protocol.
Illia Polosukhin’s Early Life and Education
Illia Polosukhin grew up in Ukraine, and by his own account, got obsessed with AI around age 10 after watching The Matrix. That early fascination shaped the trajectory of his career.
He studied Applied Mathematics and Computer Science at Kharkiv Polytechnic Institute, where he also became a serious competitive programmer, competing in ICPC, the international collegiate programming Olympics. That foundation in rigorous algorithmic thinking is what shaped him into the AI researcher he’d eventually become.
Before moving into large-scale AI research, he worked at Salford Systems on predictive analytics, which helped build his early machine learning foundation.
The “Attention Is All You Need” Moment
In 2014, Illia joined Google Research, contributing to TensorFlow, open-source machine learning tools, and natural language understanding research. His work there was significant, but 2017 became the defining moment of his AI career.
That year, he was one of eight co-authors on Attention Is All You Need, the landmark paper that introduced the Transformer architecture powering modern AI models like ChatGPT, Claude, and Gemini.
Before Transformers, AI models read language sequentially, one token at a time, with limited memory of what came before. The self-attention mechanism changed that entirely, letting models process whole sequences in parallel and understand relationships across context. It’s been cited tens of thousands of times, and the eight co-authors have gone on to collectively found some of the most valuable AI companies in the world.
I find it genuinely fascinating that someone with this level of foundational AI credibility then chose blockchain. And the reason why is more grounded than you’d expect.
How an AI Startup Problem Led Illia Polosukhin to Build NEAR
In 2017, Illia co-founded what eventually became NEAR with Alexander Skidanov. But it didn’t start as a blockchain company at all. The original idea, Near.ai, was a crowdsourced data-labeling platform for AI training.
The problem came when he tried to pay contributors across Ukraine, Poland, Russia, and China. Traditional cross-border payments were slow, expensive, and unreliable for students without formal banking. Crypto solved that cleanly.
When no existing blockchain could handle what they actually needed at scale, they built one. That’s the whole origin of the NEAR Protocol.
What Is NEAR Protocol and Why It Matters
NEAR launched on mainnet in 2020 as a Layer-1 blockchain built around scalability, low fees, and developer experience. Positioning it purely against Ethereum or Solana misses the point, though.
NEAR’s Nightshade sharding architecture processes transactions across parallel segments simultaneously, hitting high throughput with finality under 600ms and fees that average fractions of a cent. The bigger differentiator today is Chain Abstraction. Through its NEAR Intents framework, users can interact with assets across 35+ blockchains without managing separate wallets or manual bridging.
The ecosystem has raised over $540 million from investors, including a16z, Tiger Global, and Coinbase Ventures, and crossed 100 million total accounts in early 2024. By October 2025, NEAR Intents had already crossed $1.8 billion in cumulative swap volume with 120,000+ users, well before Confidential Intents even launched.
Illia’s Vision for Decentralized AI and Web3 Infrastructure
Here’s where his thinking gets really interesting. Illia’s core argument is that the current AI model is dangerously centralized. A few corporations control the most powerful models, the training pipelines, and the information layer that billions of people use every day.
His answer is decentralized AI, where the model doesn’t belong to a corporation, the data belongs to you, and the training process isn’t a black box. NEAR’s Web3 infrastructure provides the trust layer, ownership mechanisms, and programmable incentives to make that actually work.
His framing is blunt: “AI is the front end of the internet, not just for blockchain, but for everything.”
Practically, this is materializing through NEAR Intents (AI agents executing cross-chain transactions autonomously) and through confidential computing research that lets AI process private data without the node operator ever seeing it. He presented that work at NVIDIA GTC 2025, making him the only Web3 founder on stage at the world’s biggest AI computing conference.
What’s Next for Illia Polosukhin and NEAR Protocol
NEAR’s near-term focus is squarely on the agentic economy. AI agents aren’t a feature on NEAR. They’re the primary design target.
The protocol is building infrastructure for autonomous agents to own assets, transact across networks, and make decisions without human sign-off at every step. Ecosystem teams raised $146 million in external capital in 2024 alone, and over 50 AI teams were actively building on NEAR by the end of that year.
In March 2026, NEAR launched Confidential Intents, bringing private cross-chain swaps to the mainnet and sending the NEAR token up 17% in a single day. It’s the clearest signal yet that the confidential computing roadmap is shipping, not just being talked about.
Illia has also been talking publicly about new market primitives for what he calls the agentic economy, where agents handle discovery, hiring, contracting, and negotiation at a scale and speed no human system can match, which is essentially physical AI playing out in the real world. If that sounds far out, the infrastructure being built right now suggests it’s closer than most people realize.
Final Thoughts
Illia Polosukhin is one of the rare AI researchers who has been genuinely foundational in two separate technological waves. The Transformer paper he co-authored is the architecture underneath every AI tool you use today. NEAR Protocol is his bet that the next wave doesn’t just replicate the same centralized power structures in a new wrapper.
You might not buy the full decentralized AI vision yet. But it’s hard to dismiss it when it’s coming from someone who already helped build the thing it’s trying to decentralize.
Want to stay ahead of where AI and blockchain are heading? At Yaabot, we cover the latest in crypto, machine learning, and the tech developments reshaping how you work and invest.
FAQs
No. He co-authored Attention Is All You Need, which introduced Transformer architecture, the foundation used in modern AI systems like ChatGPT.
He co-authored “Attention Is All You Need” in 2017, the paper that introduced the Transformer architecture. Every major AI model you use today, from ChatGPT to Gemini, is built on that foundation.
He’s one of the few people bridging serious AI research credibility with blockchain infrastructure. NEAR’s Chain Abstraction and AI agent frameworks are pushing Web3 closer to something regular users and autonomous systems can actually use without friction.
As of March 2026, NEAR was the top performer in the CoinDesk 20 Index, up 12.4% over the weekend on the back of Confidential Intents launching. The technical roadmap is shipping, and the market is noticing.
Yes, and his take is worth noting. He argues that decentralized AI is actually easier to regulate than centralized models, because the training pipelines and governance are transparent by design rather than hidden inside a corporation.
