August 4, 2013


I recently graduated from the department of Computer Science at the University of Toronto under the excellent supervision of Geoffrey Hinton. I am now a Research Scientist at Google in the Brain team. My interests lie in Machine Learning and Bioinformatics. I am currently working on new Deep Learning models for sequences, inspired by the recent success of sequence-to-sequence.

Here are links to some selected research projects and software from my PhD and earlier (see Google scholar profile for a more detailed and recent list):

  • Kaldi Recipe for TIMIT and decoding scripts - The TIMIT Kaldi recipe I used for my thesis and related papers. Please email me for instructions on how to use this recipe.
  • My PhD thesis - Exploring Deep Learning Methods for Discovering Features in Speech Signals.
  • INTERSPEECH 2014 paper and code paper on how multiframe predictions can significantly improve performance of DNN-HMM systems.
  • Paper, and presentation slides on a method to create augmented datasets for training neural networks on audio data. In Deep Learning for Audio, Speech and Language Processing, ICML 2013
  • Paper on using an autoencoder with deformable template parts to discover features in speech. In Interspeech 2013
  • Tech report, paper, and poster (in Interspeech 2012) on work I did at Google using Deep Neural Networks for Acoustic Modeling in speech recognition systems. This technology is now in use in Google Voice Search on Android and other platforms.
  • Paper on using Transforming Autoencoders for feature discovery (NIPS 2011 Workshop on Deep Learning and Unsupervised Feature Learning)
  • Paper on discovering features in raw speech signals that can be used to achieve high speech recognition accuraccy.
  • Cryobayes: An algorithm for 3D Reconstruction of Macromolecular Structure from Electron Cryo Microscopy data. See paper in Bioinformatics.
  • Decon2LS: a software tool for automatic analysis of high resolution mass spectral data. (Reference).
  • MultiAlign: a software tool for finding mass spectral features common to multiple mass spectral analyses.
  • Wiki Link


  • Email:

    ndjaitly AT cs DOT toronto DOT edu
  • Office:

    Dept of Computer Science
    Pratt Bldg, Room 275
    6 King's College Rd
    University of Toronto
    Toronto, ON M5S 3G4, Canada
  • Mailing:

    Navdeep Jaitly
    Department of Computer Science
    University of Toronto
    3302-10 King's College Road
    Toronto, Ontario CANADA M5S 3G4*