Paper Huddle

The 8 Google Researchers Who Invented the Transformer: Where Are They Now?

Where the eight Attention Is All You Need authors are as of October 2026: Cohere, Sakana AI, Essential AI, NEAR, Inceptive, and Shazeer's move to OpenAI.

By Shivam Singh · · 5 min read

As of 12 October 2026, none of the eight authors of the Transformer paper is reported to work at Google. The last one, Noam Shazeer, announced on X in June 2026 that he would join OpenAI. Five of the eight run or co-founded their own companies, from Cohere to Sakana AI.

The paper is "Attention Is All You Need" (arXiv 1706.03762), submitted on 12 June 2017. This post covers what the authors built at the time and where each one is now.

Who were the eight authors, and who did what?

The paper's author list is Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin. A footnote reads "Equal contribution. Listing order is random." So the order carries no meaning.

The same footnote records who did what. These are the authors' own descriptions:

AuthorAffiliation on the paperContribution as stated by the authors
Jakob UszkoreitGoogle ResearchProposed replacing RNNs with self-attention and started the effort to evaluate it
Ashish VaswaniGoogle BrainWith Illia, designed and implemented the first Transformer models
Illia PolosukhinNone listed (work done at Google Research)With Ashish, designed and implemented the first Transformer models
Noam ShazeerGoogle BrainProposed scaled dot-product attention, multi-head attention and the parameter-free position representation
Niki ParmarGoogle ResearchDesigned, implemented, tuned and evaluated countless model variants
Llion JonesGoogle ResearchInitial codebase, efficient inference and visualizations
Lukasz KaiserGoogle BrainDesigned and implemented tensor2tensor, which replaced the earlier codebase
Aidan N. GomezUniversity of Toronto (work done at Google Brain)Designed and implemented tensor2tensor

What did the Transformer actually achieve in 2017?

The authors report 28.4 BLEU on the WMT 2014 English-to-German task and 41.8 BLEU on English-to-French for the big model. They say the big model trained for 3.5 days on eight NVIDIA P100 GPUs, and the base model reached 27.3 and 38.1 BLEU in about 12 hours.

The strongest baselines in the paper's table were ensembles: GNMT+RL at 26.30 (English-German) and 41.16 (English-French), and ConvS2S at 26.36 and 41.29. These are lab-reported numbers on a translation benchmark. The paper does not show that the architecture would later carry general-purpose language models.

Where is Noam Shazeer now?

Shazeer posted on X in June 2026 that he would be "joining OpenAI," according to 9to5Google. He was a Gemini co-lead and VP of Engineering at Google, a role he had held since August 2024. His post did not state his new role at OpenAI. Google told Reuters it was "grateful for Noam's meaningful contributions to Google over the years."

This was his second exit. 9to5Google notes he was CEO of Character.AI from late 2021 to 2024, when he returned to Google as part of a licensing deal with the startup.

Where are Aidan Gomez and Llion Jones now?

Both lead companies they co-founded. Gomez is co-founder and CEO of Cohere. Jones is Co-Founder and CTO of Sakana AI, which the company says was founded in July 2023 with David Ha as CEO.

Cohere's recent history, as of 12 October 2026:

Where are Ashish Vaswani and Niki Parmar now?

Vaswani is CEO of Essential AI, based in San Francisco. In December 2025 the company released Rnj-1, which it describes as an 8B model that roughly follows the open-source Gemma 3 architecture. Fortune reported in February 2025 that Vaswani and Parmar co-founded Essential AI after Parmar co-founded Adept AI.

According to Fortune, writing on 13 February 2025, "Anthropic also recently hired Niki Parmar." Fortune gives no start date or role, and her current position is not confirmed here.

Where are Jakob Uszkoreit, Illia Polosukhin and Lukasz Kaiser now?

Their paths split between biotech, blockchain and OpenAI. Each of these rests on a single secondary source.

  • Uszkoreit founded Inceptive. According to Maginative, a $100 million round in September 2023 was led by NVIDIA's NVentures and Andreessen Horowitz, tripling the valuation to over $300 million.
  • Polosukhin co-founded NEAR with Alexander Skidanov. Wikipedia says the project began as Near.ai, an AI startup, and launched its mainnet in 2020.
  • Kaiser is at OpenAI. A Teamday write-up of an interview, dated 19 December 2025, says he led the research that produced the o1 reasoning models. It calls him the only one of the eight who stayed an engineer rather than founding a startup.

What pattern does this show?

Wikipedia's article on the paper says each of the eight left Google to join other companies or found startups. The table below summarizes the destinations covered above.

AuthorWhere (as of 12 October 2026)TypeSource strength
Ashish VaswaniCEO, Essential AIStartupCompany page
Noam ShazeerJoining OpenAI (announced June 2026)LabNews report quoting his post
Niki ParmarReported at Anthropic (Fortune, February 2025); current status unconfirmedLabOne outlet
Jakob UszkoreitFounder, InceptiveStartupOne outlet
Llion JonesCo-Founder and CTO, Sakana AIStartupCompany page
Aidan GomezCo-founder and CEO, CohereStartupCompany page
Lukasz KaiserOpenAILabOne outlet
Illia PolosukhinCo-founder, NEARStartupOne outlet

The sample is eight people, so treat the pattern as an anecdote, not evidence about why researchers leave large labs. The destinations are still informative: Shazeer, Parmar and Kaiser are reported at frontier labs, while Vaswani, Jones, Gomez, Uszkoreit and Polosukhin built or co-founded companies. Shazeer and Parmar have done both.

What's still unclear?

Several pieces of this story are not settled.

  • Shazeer's new role. His post did not name one.
  • Current employment of Parmar, Uszkoreit, Kaiser and Polosukhin. The sources are single outlets, some dated well before October 2026. Titles and employers may have changed.
  • Credit. The contribution note is the authors' own account. Individual credit for the ideas has been described in different ways elsewhere, and this post does not rank the contributions.

Frequently asked questions

Who wrote Attention Is All You Need?

Eight people: Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin. The arXiv paper (1706.03762) was submitted on 12 June 2017. A footnote marks all eight as equal contributors and says the listing order is random, so there is no first author.

Do any of the Transformer authors still work at Google?

Not as of 12 October 2026, as far as public reports show. Wikipedia says each of the eight left to join other companies or found startups. The last holdout, Noam Shazeer, announced on X in June 2026 that he would join OpenAI, according to 9to5Google.

Which Transformer authors founded companies?

Aidan Gomez co-founded Cohere, Llion Jones co-founded Sakana AI, Ashish Vaswani leads Essential AI (co-founded with Niki Parmar), Jakob Uszkoreit founded Inceptive, and Illia Polosukhin co-founded NEAR. Shazeer co-founded Character.AI before returning to Google in 2024.

How good were the original Transformer's results?

The authors report 28.4 BLEU on WMT 2014 English-to-German and 41.8 on English-to-French for the big model. They trained it for 3.5 days on one machine with eight NVIDIA P100 GPUs. These are the authors' own results on a translation benchmark.

Who proposed self-attention in the Transformer paper?

The paper's contribution note says Jakob Uszkoreit proposed replacing RNNs with self-attention. It credits Vaswani and Polosukhin with the first Transformer models, and Shazeer with scaled dot-product attention, multi-head attention and the parameter-free position representation.

Sources

  1. Attention Is All You Need (arXiv abstract page), arXiv (2023-08-02)
  2. Attention Is All You Need (arXiv HTML, v7), arXiv (2023-08-02)
  3. Gemini's co-lead is leaving Google to join OpenAI, 9to5Google (2026-06-17)
  4. Where have the eight Transformer pioneers ended up, BestHub (repost of Machine Heart) (2026-06-28)
  5. Sakana AI company info, Sakana AI (2026-10-12)
  6. About Cohere, Cohere (2026-10-12)
  7. Cohere and Aleph Alpha sign agreement, Cohere (2026-09-16)
  8. Cohere secures $500 million USD at $6.8-billion valuation, BetaKit (2025-08-14)
  9. Essential AI about page, Essential AI (2026-10-12)
  10. Rnj-1 announcement, Essential AI (2025-12-05)
  11. Anthropic hired Daniela Amodei's husband, Fortune (2025-02-13)
  12. Inceptive Raises $100 Million to Design New Vaccines and Therapies with AI, Maginative (2023-09-07)
  13. NEAR (blockchain platform), Wikipedia (2026-10-12)
  14. Lukasz Kaiser: Reasoning Models Are Just Getting Started, Teamday (2025-12-19)
  15. Attention Is All You Need (Wikipedia), Wikipedia (2026-10-09)