Slav Petrov
Also known as Slav Orlinov Petrov
Slav Petrov is a natural language processing researcher who helped build tools for multilingual parsing and contributed to large language models. He earned a UC Berkeley PhD in 2009, helped develop shared part-of-speech and dependency resources, and coauthored work on SyntaxNet and PaLM. The ACL recognized his work with a 2014 John Atanasoff Award and a 2023 Test-of-Time Award for his 2013 dependency-annotation paper. As of October 2026, he is vice president of research at Google DeepMind and co-leads Gemini post-training.
Timeline
-
2009
A Berkeley NLP doctorate
He completes his PhD at UC Berkeley in 2009 with the dissertation Coarse-to-Fine Natural Language Processing.
-
2012
Universal part-of-speech tags
In 2012, he coauthors A Universal Part-of-Speech Tagset with Dipanjan Das and Ryan McDonald; the paper maps treebank labels into a shared set used across 22 languages.
-
2013
Multilingual dependency annotation
In 2013, he coauthors Universal Dependency Annotation for Multilingual Parsing with twelve other researchers.
-
2014
A Bulgarian research award
He receives the 2014 John Atanasoff Award from Bulgaria's president.
-
2017
A baseline parser for CoNLL
In 2017, he coauthors the CoNLL shared-task report on ParseySaurus, a dependency parser that beats Parsey's Cousins by 3.47% absolute labeled accuracy across 52 Universal Dependencies treebanks.
-
2022
PaLM and Pathways
In 2022, he coauthors the PaLM paper, which describes a 540-billion-parameter model trained with Pathways and reports results on language, reasoning, code, and multilingual tasks.
-
2023
An ACL Test-of-Time Award
The ACL gives its 10-Year Test of Time Paper Award to his 2013 coauthored paper in 2023.
-
As of October 2026
Leading Gemini post-training
As of October 2026, he is vice president of research at Google DeepMind and co-leads post-training for Gemini across audio, multilinguality, and tool use.
Sources and credits
Photo: We have no published photo of Slav Petrov yet. Send us a free-licensed photo.
- Ph.D. Dissertations | EECS at UC Berkeley
- Coarse-to-Fine Natural Language Processing
- Slav Petrov
- A Universal Part-of-Speech Tagset - ACL Anthology
- Universal Dependency Annotation for Multilingual Parsing - ACL Anthology
- SyntaxNet Models for the CoNLL 2017 Shared Task
- PaLM: Scaling Language Modeling with Pathways
- Köpfe des AI Centers