Person

Denis Yarats

Denis Yarats is a computer scientist and machine-learning researcher whose work includes reinforcement learning; he co-founded Perplexity and serves as its chief technology officer.[1]first-party sourcePerplexity Sonar Dominates New Search Arena EvaluationPerplexityPerplexity identifies Yarats as a co-founder and CTO; his research role and technical work are documented in his NYU-affiliated papers.CheckedOpen original source

Last reviewedMarkdown

On this page

01

Core facts

Entity type
Person
Employer
Perplexity AI[1]first-party sourcePerplexity Sonar Dominates New Search Arena EvaluationPerplexityPerplexity identifies Yarats as its co-founder and CTO.CheckedOpen original source
Occupations
Computer scientist; machine-learning researcher; technology entrepreneur[2]first-party sourceNYU Entrepreneurs Festival — 2025 agendaNYU EntrepreneurshipNYU identifies Yarats as a Courant PhD alumnus and Perplexity co-founder and CTO; his conference papers document his computer-science research.CheckedOpen original source

02

Who is Denis Yarats?[3]independent sourceAI-powered search engine Perplexity AI lands $26M, launches iOS appTechCrunchTechCrunch documents Perplexity’s 2022 founding and Yarats’s Quora and Meta AI career; NYU documents his PhD and Perplexity role; his research papers describe his reinforcement-learning contributions.CheckedOpen original source [2]first-party sourceNYU Entrepreneurs Festival — 2025 agendaNYU EntrepreneurshipNYU identifies Yarats as a PhD (Courant ’22) and Perplexity co-founder and CTO; his cited papers document research in visual reinforcement learning.CheckedOpen original source

Yarats’s research focuses on reinforcement learning, including methods for training agents directly from visual observations. He completed a computer-science PhD at New York University’s Courant Institute in 2022. Earlier in his career, he worked at Yandex and Microsoft, then led machine-learning work at Quora before joining Facebook AI Research. His research papers include work on sample-efficient visual control, data-regularized Q-learning (DrQ), and DrQ-v2. In 2022, he co-founded Perplexity with Aravind Srinivas, Johnny Ho, and Andy Konwinski; the company identifies him as its co-founder and CTO.

03

History

  1. Co-authored research on hierarchical text generation and planning.[4]independent sourceHierarchical Text Generation and Planning for Strategic DialogueProceedings of Machine Learning ResearchThe PMLR record lists Denis Yarats as an author of the 2018 paper on hierarchical text generation, planning and reinforcement learning.CheckedOpen original source

  2. Co-authored research on reinforcement learning with prototypical representations.[5]independent sourceReinforcement Learning with Prototypical RepresentationsProceedings of Machine Learning ResearchThe PMLR record lists Denis Yarats as an author of this 2021 reinforcement-learning paper.CheckedOpen original source

  3. Perplexity was founded in 2022 by a team that included Denis Yarats.[6]independent sourceInside the buzzy AI startup coming for Google’s lunchForbesForbes reports that Yarats and the other co-founders decided to build the application in 2022.CheckedOpen original source

  4. Co-founded Perplexity in 2022 with Aravind Srinivas, Johnny Ho, and Andy Konwinski.[3]independent sourceAI-powered search engine Perplexity AI lands $26M, launches iOS appTechCrunchTechCrunch’s April 2023 report names all four founders and dates the company’s founding to 2022.CheckedOpen original source

04

Education

Belarusian State University
Studied applied mathematics and information technology at Belarusian State University from 2005 to 2011.[7]independent sourceDenis Yarats on machine learning, reinforcement learning and his careerdev.byIn a 2019 interview, Yarats described studying at Belarusian State University’s Faculty of Applied Mathematics and Computer Science from 2005 to 2011.CheckedOpen original source
Nyu Phd
Completed a PhD in computer science at New York University’s Courant Institute in 2022.[2]first-party sourceNYU Entrepreneurs Festival — 2025 agendaNYU EntrepreneurshipNYU’s 2025 festival agenda identifies Denis Yarats as PhD (Courant ’22).CheckedOpen original source

05

Career and roles

Meta Fair
Joined Facebook AI Research (FAIR) in 2016 as a research engineer and worked on machine learning and reinforcement learning.[7]independent sourceDenis Yarats on machine learning, reinforcement learning and his careerdev.byThe 2019 interview identifies him as a Facebook Research Engineer working at FAIR and dates the beginning of that role to 2016.CheckedOpen original source
Microsoft
Worked at Microsoft in Seattle from 2011 to 2013 on maps and the Bing search engine.[7]independent sourceDenis Yarats on machine learning, reinforcement learning and his careerdev.byYarats’s 2019 interview dates his Microsoft role to 2011–2013 and describes work on maps and Bing.CheckedOpen original source
Quora
Worked at Quora from 2013 to 2016, where he led machine-learning work on news-feed ranking and email digests.[7]independent sourceDenis Yarats on machine learning, reinforcement learning and his careerdev.byYarats’s 2019 interview describes his Quora role, dates it to 2013–2016, and names the news-feed ranking and email-digest systems.CheckedOpen original source
Yandex
Worked at Yandex in 2011 and served as a curator for its School of Data Analysis in Minsk.[7]independent sourceDenis Yarats on machine learning, reinforcement learning and his careerdev.byYarats’s 2019 interview recounts his 2011 Yandex role and work as a curator with the Minsk School of Data Analysis.CheckedOpen original source

06

Contributions and work

Drq
The DrQ project released code and experimental configurations alongside its paper.[8]first-party sourceDrQ research implementationDenis YaratsYarats’s public research repository provides the DrQ implementation and reproduction instructions.CheckedOpen original source
Drqv2
The DrQ-v2 research team published an implementation for reproducing the visual-control experiments.[9]first-party sourceDrQ-v2 research implementationFacebook ResearchThe Facebook Research repository describes itself as the DrQ-v2 implementation and links to the paper.CheckedOpen original source
Drq
Co-developed DrQ (Data-regularized Q), which applies image augmentation and Q-value regularisation to model-free reinforcement learning.[10]structured public data sourceImage Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from PixelsarXivThe paper introduces DrQ and explains its data-regularisation approach for learning from image observations.CheckedOpen original source
Drq Equal Contribution
Was a co-first author, with Ilya Kostrikov, of the DrQ paper; the authors state that author order was determined by coin flip.[11]structured public data sourceImage Augmentation Is All You Need — ICLR 2021OpenReview / ICLRThe ICLR 2021 paper marks Yarats and Kostrikov with equal-contribution asterisks and states that author order was determined by coin flip.CheckedOpen original source
Drqv2
Co-authored DrQ-v2, a model-free reinforcement-learning method for visual continuous control.[12]structured public data sourceMastering Visual Continuous Control: Improved Data-Augmented Reinforcement LearningarXivThe paper lists Yarats among its authors and describes DrQ-v2 as a data-augmented method for visual continuous control.CheckedOpen original source
Prototypical Representations
Co-authored a 2021 paper on reinforcement learning with prototypical representations.[13]structured public data sourceReinforcement Learning with Prototypical RepresentationsarXivThe arXiv record lists Yarats, Rob Fergus, Alessandro Lazaric, and Lerrel Pinto as authors.CheckedOpen original source
Reinforcement Learning Focus
His published work includes reinforcement learning from images, sample-efficient visual control, and strategic dialogue generation.[14]first-party sourceDenis Yarats — NYU Computer ScienceNew York UniversityYarats’s NYU research page and linked publication record cover visual reinforcement learning and language-generation research.CheckedOpen original source
Sample Efficient Rl
Co-authored research on improving sample efficiency in model-free reinforcement learning from images.[15]structured public data sourceImproving Sample Efficiency in Model-Free Reinforcement Learning from ImagesarXivThe arXiv paper lists Yarats among its authors and presents methods for training reinforcement-learning agents from image observations.CheckedOpen original source
Strategic Dialogue
Co-authored research on hierarchical text generation and planning for strategic dialogue.[16]structured public data sourceHierarchical Text Generation and Planning for Strategic DialoguearXivThe arXiv record lists Yarats among the authors of the paper on hierarchical text generation and planning.CheckedOpen original source

07

Leadership

Perplexity Cto
Serves as co-founder and chief technology officer of Perplexity.[1]first-party sourcePerplexity Sonar Dominates New Search Arena EvaluationPerplexityPerplexity’s 2025 API-platform event page identifies Yarats as its co-founder and CTO.CheckedOpen original source
Role
Co-founder and chief technology officer of Perplexity[1]first-party sourcePerplexity Sonar Dominates New Search Arena EvaluationPerplexityPerplexity identifies Denis Yarats as its co-founder and CTO in its official API announcement.CheckedOpen original source

09

Sources

Reference numbers stay stable across this page and each source appears once, ordered by first visible use.

  1. NYU Entrepreneurs Festival — 2025 agenda

    NYU Entrepreneurship · first-party source

  2. Hierarchical Text Generation and Planning for Strategic Dialogue

    Proceedings of Machine Learning Research · independent source

  3. Reinforcement Learning with Prototypical Representations

    Proceedings of Machine Learning Research · independent source

  4. DrQ research implementation

    Denis Yarats · first-party source

  5. DrQ-v2 research implementation

    Facebook Research · first-party source

  6. Image Augmentation Is All You Need — ICLR 2021

    OpenReview / ICLR · structured public data source

  7. Denis Yarats — NYU Computer Science

    New York University · first-party source

  8. timesofindia.indiatimes.com

    timesofindia.indiatimes.com · first-party source

Found something that needs review?

Suggest a correction