AI People Wiki

Builders & open source

Michael Auli

Former Principal Research Scientist/Director at Meta FAIR; last author of wav2vec 2.0, fairseq, and MMS

Michael Auli's homepage, last updated in April 2024, describes Auli as a Principal Research Scientist/Director at Meta's FAIR lab working on machine learning, speech processing, and natural language processing, with projects such as wav2vec, the fairseq toolkit, and convolutional sequence-to-sequence models; the 2025 Omnilingual ASR paper marks Auli's contribution as work conducted while at FAIR at Meta. Before Meta, Auli was at Microsoft Research, working on early neural machine translation and neural dialogue models. Auli is the last author of the wav2vec 2.0, data2vec, and fairseq papers and of Scaling Speech Technology to 1,000+ Languages, the Massively Multilingual Speech (MMS) project.1234567

At a glance

Current role
None confirmed in the sources
Education
Edinburgh
Work listed
3
Sources cited
12
Profile reviewed

Career

Auli's doctoral work at the University of Edinburgh, advised by Adam Lopez and Philipp Koehn, was on CCG parsing. At Microsoft Research, Auli was involved in early work on neural machine translation and neural dialogue models, including a 2015 paper on generating context-sensitive conversational responses with neural networks.128

At Facebook AI Research, Auli led the team that developed convolutional sequence-to-sequence models and later worked as a Principal Research Scientist/Director on self-supervised learning for speech and language. The 2025 Omnilingual ASR paper marks Auli's contribution as work conducted while at FAIR at Meta.213

Research

The 2017 convolutional sequence-to-sequence paper replaced recurrent networks with convolutions for sequence learning. Auli then co-authored the wav2vec line of self-supervised speech models: wav2vec (2019) pre-trained on unlabeled audio for speech recognition, and wav2vec 2.0 (2020) learned speech representations by masking latent audio features.9104

data2vec (2022) applied a single self-supervised method to speech, vision, and language by predicting contextualized latent representations, and data2vec 2.0 made that training more efficient. The Massively Multilingual Speech project (2023) built pre-trained wav2vec 2.0 models covering 1,406 languages and a speech recognition model for 1,107 languages, and Omnilingual ASR (2025) extended open multilingual speech recognition to more than 1,600 languages.511713

Open-source work

Auli is the last author of the 2019 paper describing fairseq, Facebook AI Research's MIT-licensed PyTorch toolkit for sequence modeling, which Auli's homepage lists among the projects from this work.6121

Work and contributions

  • fairseq Open-source code

    Last author of fairseq: A Fast, Extensible Toolkit for Sequence Modeling (2019); Auli's homepage lists the fairseq toolkit among the projects from this work, and the repository is published under the MIT license.6112

  • wav2vec 2.0 Published research

    Last author of wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations (2020), with Alexei Baevski as first author.41

  • Massively Multilingual Speech (MMS) Published research

    Last author of Scaling Speech Technology to 1,000+ Languages (2023), which describes pre-trained wav2vec 2.0 models covering 1,406 languages and a multilingual speech recognition model for 1,107 languages.71

Timeline

  1. Co-authored a paper on neural-network generation of context-sensitive conversational responses.8
  2. Co-authored "Convolutional Sequence to Sequence Learning" at Facebook AI Research.9
  3. Released the fairseq toolkit paper as last author.6
  4. Co-authored wav2vec, on unsupervised pre-training for speech recognition.10
  5. Released wav2vec 2.0 as last author, with Alexei Baevski as first author.4
  6. Released data2vec, a general self-supervised learning framework for speech, vision, and language.5
  7. Released data2vec 2.0, a more efficient version of the data2vec learning objective.11
  8. Released the Massively Multilingual Speech (MMS) paper as last author.7
  9. Omnilingual ASR was released, with Auli's contribution marked as work conducted while at FAIR.3

Documented roles

The sources do not confirm a current role; these are the roles they document.

Companies in their history

Suggest a correction or removal

Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.

References

12 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026

  1. Michael Auli — homepageMichael Auli · Official page · accessed Oct 21, 2026
  2. Michael Auli — AI at Meta people pageMeta · Official page · accessed Oct 21, 2026
  3. Omnilingual ASR: Open-Source Multilingual Speech Recognition for 1600+ Languages — paper PDF (Auli marked: work conducted while at FAIR at Meta)arXiv · Paper · published Nov 12, 2025 · accessed Oct 21, 2026
  4. wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations (arXiv 2006.11477)arXiv · Paper · published Jun 20, 2020 · accessed Oct 21, 2026
  5. data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language (arXiv 2202.03555)arXiv · Paper · published Feb 7, 2022 · accessed Oct 21, 2026
  6. fairseq: A Fast, Extensible Toolkit for Sequence Modeling (arXiv 1904.01038)arXiv · Paper · published Apr 1, 2019 · accessed Oct 21, 2026
  7. Scaling Speech Technology to 1,000+ Languages (arXiv 2305.13516)arXiv · Paper · published May 22, 2023 · accessed Oct 21, 2026
  8. A Neural Network Approach to Context-Sensitive Generation of Conversational Responses (arXiv 1506.06714)arXiv · Paper · published Jun 22, 2015 · accessed Oct 21, 2026
  9. Convolutional Sequence to Sequence Learning (arXiv 1705.03122)arXiv · Paper · published May 8, 2017 · accessed Oct 21, 2026
  10. wav2vec: Unsupervised Pre-training for Speech Recognition (arXiv 1904.05862)arXiv · Paper · published Apr 11, 2019 · accessed Oct 21, 2026
  11. Efficient Self-supervised Learning with Contextualized Target Representations for Vision, Speech and Language (arXiv 2212.07525)arXiv · Paper · published Dec 14, 2022 · accessed Oct 21, 2026
  12. facebookresearch/fairseq README and licenseMeta (GitHub repository) · Repository · accessed Oct 21, 2026