robfelty.com

Résumé

Robert Felty
52146 Würselen, Germany
robfelty@gmail.com

Technical Skills

  • Programming languages — Python, Java, groovy, PHP, Perl, R, and BASH

  • Natural Language Processing — Information Retrieval (lucene, solr, elasticsearch), supervised classification (log-linear models, SVM), named entity recognition and parsing (HMM, CRF, PCFG)

  • Deep Learning — TensorFlow, Theano, Keras, BLSTM, seq2seq, word2vec

  • Automated Speech Recognition VoiceXML, ECMAscript, SRGS and JSGF grammars

  • Acoustic analysis — spectral and formant analysis using Praat, Matlab, and R

  • Big Data — Hadoop architecture (Cloudera Manager), ETL ( Kafka, Flume), analytics (Hive, Impala, Spark)

  • Statistics — t-test, correlations, ANOVA, multiple regression, logistic regression, monte-carlo, and bootstrapping, using python and R

  • Web development — html, css, javascript, jQuery, WordPress, MySQL, PostgreSQL, PHP, XML

  • Source content management — git, Subversion, mercurial

Experience

  • Nuance Communications (renamed to Cerence Oct. 2019)

    • Principal Research Scientist — 2016–present

      • Implementing new features and improved accuracy for natural language understanding for German localization of automotive virtual assistant

      • Developing internal tools for accuracy analysis and reporting in python and groovy

  • Automattic

    • Data Engineer / Scientist — 2015–2016

      • Created tools for better monitoring and alerting of Hadoop ecosystem

      • Developed complex queries for business analytics

      • Built visualization tools and dashboards for internal customers

      • Led team of 3 developers

  • Nuance Communications

    • Senior Research Engineer — 2014–2015

      • Technical lead for project to create media knowledge base to support personal assistants

      • Created solr indices from a collection of sources to support searching of music, movies, and tv shows for 13 locales

      • Improved search relevancy by performing experiments and developing tools to tune Lucene parameters.

    • Research Engineer — 2011–2014

      • Technical lead developing natural language understanding for tv-based dialogue system, using a UIMA-based pipeline with a combination of statistical and grammar-based components for 8 languages

      • Performed experiments for measuring and improving accuracy on millions of utterances using sun grid engine

      • Provided analytics for usage data using a combination of BASH, python, and R

    • Speech Scientist — 2010–2011

      • Created SRGS grammars for automated telephony systems

      • Performed speech recognition experiments to improve performance of enterprise intelligent interactive voice response applications

      • Developed and improved internal tools to improve workflow efficiency in BASH, Perl, python, and Java

  • Lecturer in Linguistics — University of Colorado – Boulder, 2009

    • Taught Introduction to Computational Corpus Linguistics

    • Topics include regular expressions, part of speech tagging, and word frequency analyses using python

  • Adjunct professor of Linguistics — Indiana University, 2008

    • Taught Introduction to programming for computational linguists

    • Topics included UNIX basics, regular expressions, searching corpora and databases, using python

  • Post-doctoral fellow — Indiana University, 2007–2009

    • Carried out a large-scale investigation of spoken word recognition of English.

    • Designed and implemented novel error analyses using perl, python, and R, including monte-carlo simulations of behavioral data.

    • Developed a new clinical screening test of speech recognition for cochlear implant candidates

  • Phonetics lab research assistant — University of Michigan, 2004–2005

    • Assisted on Dr. José Benkí’s project Measuring context effects in spoken word recognition.

    • Managed and mentored four undergraduate research assistants

    • Performed acoustic analysis and editing of stimuli for speech perception experiments

    • Created custom stimulus presentation software and data analysis software using Matlab

Education

University of Michigan

  • Ph.D. in Linguistics and Germanic Languages & Literatures — May, 2007

    • Dissertation: Context Effects in Spoken Word Recognition of English and German by Native and Non-native Listeners

    • taught five undergraduate courses, including the first four semesters of German, and Introduction to Linguistics

  • M.A. in Linguistics — January 2005

    • Thesis: The Ongoing (Un)merging of Stops in the Swabian Dialect of Isny im Allgäu

Grinnell College

  • B.A. in German, with honors — December, 2000

  • Semester Abroad in Munich, Germany — Spring, 1999

Online Courses

  • Mining Massive Datasets— Coursera, from Jeff Ulman, Stanford University — November, 2014

  • Machine Learning — Coursera, from Andrew Ng, Stanford University — May, 2014

  • Natural Language Processing — Coursera, from Michael Collins, Columbia University — May, 2013

Selected Publications

Languages

  • Fluent in English and German

  • Linguistic knowledge of Spanish (particular phonetic and phonological), as well as rudimentary speaking skills

Awards

  • Nuance Chairman’s club recipient — 2012

  • Candidacy Fellowship (awarded by the Linguistics department) — 2005

  • Regents Fellowship from the University of Michigan — 2001

  • Phi Beta Kappa Member — 2001

  • Trustee Merit scholarship from Grinnell College — 1997–2000