Ashish Vaswani is a computer scientist and the lead author of "Attention Is All You Need" (2017), the paper written with seven co-authors at Google that introduced the Transformer (architecture) architecture underlying nearly all modern large language models.
The transformer paper
Vaswani holds a PhD in computer science from the University of Southern California, where he worked on statistical machine translation before joining Google Brain. As lead author of the transformer paper, he was central to proposing that a sequence transduction model could dispense with recurrence and convolutions entirely and rely solely on a self-attention mechanism to model relationships between all positions in a sequence in parallel. The resulting architecture proved dramatically more efficient to train at scale than the recurrent networks it replaced, and it directly enabled the pretraining of much larger models, forming the basis for BERT, the GPT series, and effectively every major model released by OpenAI, Anthropic, and Google DeepMind since.
After Google
Vaswani left Google in 2020. In 2022, he co-founded Adept AI alongside fellow transformer co-author Niki Parmar and other former Google and OpenAI researchers, a startup focused on building AI agents capable of taking actions inside software interfaces, an early entrant into what became the broader AI agent and Computer use (AI) wave of 2024 to 2026. Adept's technology and a portion of its team were absorbed by Amazon in a 2024 licensing and talent deal, a structure similar to those used in other large labs' acquisitions of smaller AI startups.
Vaswani subsequently co-founded Essential AI, another startup focused on foundation model research for enterprise applications, continuing to work on the kind of large-scale Pretraining research that descended directly from his original transformer work.
Influence
As the lead author of the single most cited paper in the field's recent history, Vaswani is widely regarded as one of the architects of the entire modern generative AI industry. The paper's eight authors, including Noam Shazeer and Aidan Gomez, all eventually left Google, and their subsequent, dispersed careers across startups and rival labs have often been cited as emblematic of how concentrated and mobile transformer-era AI talent became after 2017.