Kelvin Guu
Stanford AI Ph.D. candidate

About me

Profile picture of Kelvin Guu

I’m a 6th-year Ph.D. student in artificial intelligence at Stanford advised by professor Percy Liang in the Stanford Natural Language Processing (NLP) Group. I completed my undergraduate degree in mathematics at Duke University, where I was advised by professors David Dunson and Alex Hartemink.

In our research, my collaborators and I have developed new methods for text generation, program synthesis from natural language, and knowledge base completion.

Long-term, I’m excited about leveraging technology to help people with high-stakes information needs (e.g. a patient trying to diagnose their rare disease, or a legislator trying to predict the impact of a new policy).

  • Areas of interest: artificial intelligence, machine learning, deep learning, natural language processing, semantic parsing, reinforcement learning, statistics

  • Applications of interest: voice user interfaces, natural language interfaces, AI-human hybrid systems, AI in healthcare, recommender systems

  • Fellowships:

    • NSF Graduate Research Fellowship (2012-2015)
    • Stanford Math+X Fellowship (2012-2013)
    • Greylock X Fellows Program (2017)

Timeline

Education

2012 - Present Ph.D. Candidate in Statistics at Stanford University
  Advisor: Percy Liang. Ph.D. Committee: Percy Liang, Wing Hung Wong, Chris Manning, and Lester Mackey.
2007 - 2011 B.S. in Mathematics at Duke University

Work experience

Summer 2015 Ph.D. Software Engineering Intern at Google
  Worked with Senior Staff Software Engineer Jakob Uszkoreit on applied natural language processing projects.
Winter 2014/15 Research Consultant at MetaMind
  Deep learning startup acquired by Salesforce in 2016. Worked with Founder & CEO/CTO Richard Socher.
2011 - 2012 Researcher in Bayesian Statistics at Duke University
  Research with Professors David B. Dunson and Alex Hartemink.
Summer 2010 Software Engineering and Product Management Intern at Redwood Systems
  Cleantech lighting automation startup founded by ex-Cisco executives and acquired in 2013 by CommScope.

Publications

  • Reinforcement Learning on Web Interfaces Using Workflow-Guided Exploration
    Kelvin Guu*, Evan Z. Liu*, Panupong Pasupat*, Tianlin Shi, Percy Liang
    International Conference on Learning Representations (ICLR), 2018
    (* equal contribution)

    pdf arxiv github bibtex

    pic

    @inproceedings{guu2018web,
      title = {Reinforcement Learning on Web Interfaces Using Workflow-Guided Exploration},
      author = {Guu, Kelvin and Liu, Evan Z. and Pasupat, Panupong and Shi, Tianlin and Liang, Percy},
      booktitle = {International Conference on Learning Representations (ICLR)},
      year = {2018},
    }
    
  • Generating Sentences by Editing Prototypes
    Kelvin Guu*, Tatsunori B. Hashimoto*, Yonatan Oren, Percy Liang
    Transactions of the Association for Computational Linguistics (TACL), 2018
    (* equal contribution)

    pdf arxiv github codalab bibtex

    pic

    @inproceedings{guu2018generating,
      title = {Generating Sentences by Editing Prototypes},
      author = {Guu, Kelvin and Hashimoto, Tatsunori B. and Oren, Yonatan and Liang, Percy},
      booktitle = {Transactions of the Association for Computational Linguistics (TACL)},
      year = {2018},
    }
    
  • From Language to Programs: Bridging Reinforcement Learning and Maximum Marginal Likelihood
    Kelvin Guu, Panupong Pasupat, Evan Liu, Percy Liang
    Association for Computational Linguistics (ACL), 2017

    pdf arxiv slides github codalab bibtex

    pic

    @inproceedings{guu2017bridging,
      title = {From Language to Programs: Bridging Reinforcement Learning and Maximum Marginal Likelihood},
      author = {Guu, Kelvin and Pasupat, Panupong and Liu, Evan Zheran and Liang, Percy},
      booktitle = {Association for Computational Linguistics (ACL)},
      year = {2017},
    }
    
  • Best paper honorable mention
    Traversing Knowledge Graphs in Vector Space
    Kelvin Guu, John Miller, Percy Liang
    Empirical Methods in Natural Language Processing (EMNLP), 2015

    pdf arxiv slides codalab bibtex

    pic

    @inproceedings{guu2015traversing,
      title = {Traversing Knowledge Graphs in Vector Space},
      author = {Guu, Kelvin and Miller, John and Liang, Percy},
      booktitle = {Empirical Methods in Natural Language Processing (EMNLP)},
      year = {2015},
    }
    
  • Bayesian Multiscale Modeling of Closed Curves in Point Clouds
    Kelvin Gu1, Debdeep Pati, David B. Dunson
    Journal of the American Statistical Association Vol. 109 (508), 2014

    pdf ncbi

    pic

  • Light timeout optimization
    Xin Gu, Kelvin Gu1, Deepak Nulu
    Redwood Systems
    US Patent No. 8,538,596, 2013

    pdf google patents

    pic

Other projects

Presentations and tutorials

Libraries

Unfortunately, much of my recent code is closed-source at the moment, but I’m working on getting it cleaned up and ready for others to use! The following projects are fairly old/out-of-date, but I’m keeping them around in the odd chance they might be of use to someone somewhere.

  • Simple Speech Recognition - 2013
    A complete speech recognition system you can deploy with just a few lines of Python, built on CMU Sphinx-4.

  • Gitmemoizer - 2013
    Automagically memoize any function and save results to disk (for Java).

  • Giterable - 2013
    A Java package for iterating through files in your Git repo and loading them.

Collaborators

I have had the privilege of working with and learning from great mentors and collaborators, including:

I’ve enjoyed mentoring younger students, including:

  • John Miller, undergraduate in Computer Science at Stanford, currently EECS PhD student at UC Berkeley
  • Justin Fu, MS in Computer Science at Stanford, currently EECS PhD student at UC Berkeley
  • Dora Demszky, undergraduate in linguistics at Princeton, currently Linguistics PhD student at Stanford
  • Evan Liu, undergraduate and master’s student in Computer Science at Stanford

Finally, I think the following entities are pretty swell:

Contact

  1. A point of clarification about my name. In the past, I’ve published under the name “Kelvin Gu”, but I now go by Kelvin Guu with two “u”s. Sorry for the confusion!  2