driving a boat in the red sea
Bridger (Waleed) Ammar
wammar atttttt holistic-intelligence.net

I'm working with a world class team to build a holistic intelligence platform for experiential learning. In 2023, I rejoined the Allen Institute for Artificial Intelligence (AI2) to help launch the Scientific Literature Graph as a standalone product. As the product manager responsible for the APIs, I wore a multitude of hats, including quarterly OKR planning, prioritizing tasks, communicating our plans to the scientific advisory board, engaging the scientific community, among others. In the past, I led the Semantic Scholar research team's early efforts to develop innovative AI-based tools to facilitate access to the scientific literature such as the literature graph [NAACL] and supp.ai [ACL Demo]. I occasionally teach courses at UW linguistics as an affiliate faculty member. As a senior research scientist at Google, I developed semantic parsing models for Google Assistant and helped develop transformer-based models for generating DNA sequences based on PacBio long-reads which significantly reduces variant-calling errors and improves assembly base accuracy [Nature Biotechnology]. In 2016, I received a Ph.D. degree in artificial intelligence from Carnegie Mellon University. Before pursuing the Ph.D., I was a research engineer at Microsoft Research, web developer at eSpace Technologies, and teaching assistant at Alexandria University.

Publications [semantic scholar, gscholar]

Professional Experience

Academic Services


Activities Log

  • Oren Etzioni --Allen Institute for Artificial Intelligence (ex-manager)
  • Noah A. Smith --University of Washington (PhD advisor)
  • Chris Dyer --Carnegie Mellon University (PhD advisor)
  • Tom Mitchell --Carnegie Mellon University (PhD thesis committee)
  • Kuzman Ganchev --Google Research (PhD thesis committee)
  • Miguel Ballesteros --Carnegie Mellon University (co-author, collaborator on cross-lingual parsing)
  • D. Sculley --Google Research (internship host)
  • Kristina Toutanova --Microsoft Research (co-author, internship host)
  • Kareem Darwish --Qatar Computing Research Institute (co-author, ex-manager)
  • Ayman Kaheel --Yahoo Inc. (ex-manager)
  • Tarek Elabbady --Microsoft Research (ex-manager)
  • Mei-Yuh Hwang --Microsoft Research (ex-manager)
  • Yulia Tsvetkov --Carnegie Mellon University (co-author, colleague)
  • Ahmed Hefny --Carnegie Mellon University (co-author, colleague)
  • Ali ElKahki --Google Research (ex-colleague at MSR Cairo)
  • Chu-Cheng Lin --Carnegie Mellon University (co-author, collaborator on modeling code switching with CRF autoencoders)
  • Phoebe Mulcaire --University of Washington (collaborator on estimating multilingual word embeddings)
  • Pradeep Dasigi --Carnegie Mellon University (collaborator on modeling selectional preferences with CRF autoencoders)
  • Moustafa Youssef --Egypt-Japan University of Science and Technology (co-author, M.Sc. ex-advisor)
  • Jeffrey Micher --US Army Research Lab (collaborator on the low-density MT project)
  • Norman Sadeh --Carnegie Mellon University (lead principal investigator of the usable privacy policy project)
  • George Foster --National Research Council Canada (Google fellowship research mentor)
  • Lori Levin --Carnegie Mellon University (co-author)
  • Jaime Carbonell --Carnegie Mellon University (department head, lead principal investigator of the low-density MT project)

Recent Projects
  • Language-universal dependency parsing* (code).
  • CRF autoencoder models for Scalable and feature-rich unsupervised learning* (code).
  • Multilingual word embeddings (unification-based*).
  • A universal depenency treebanks analyzer* (code).
  • Large-scale online training of random forests.*
  • Bayesian models for record linkage* (code).
  • CRF model for transliteration* (code).
  • Dual decomposition of a CFG parser and a POS tagger* (code)
  • A bunch of handy C, C++ and python utilities* (code).
  • Privacy policy crawler* (code).
  • C++ library for training recurrent neural network (code).
  • A neural network model which generalizes CRF autoencoders, for modeling selectional preferences (code).
  • A computational model for linguistic borrowing (code).
  • Semi-supervised learning for token-level language identification. (task, Twitter results, surprise genre results)
  • Improved training and model selection of unsupervised sequence-labeling models with lexical constraints.
  • Yet another implementation of the dependency parsing with DMV* (code).
  • Yet another implementation of logistic regression* (code).
  • Yet another implementation of word-alignment induced preordering for machine translation* (code).
Projects led by me are marked with *