PeopleofSuperintelligence

Researcher

Dan Ellis

Also known as Daniel P. W. Ellis, Daniel Ellis

Dan Ellis is a machine-listening researcher whose work helps computers recognize speech, music and environmental sounds. He grew up in England, studied engineering at Cambridge, earned an MIT doctorate in 1996 and led Columbia's LabROSA from 2000 to 2015 before joining Google. He announced AudioSet and coauthored research on stronger sound-event labels and MuLan's music-text model. IEEE named him a Fellow in 2018, and a 2023 Google application named him as an inventor on structured sound records. In July 2025, he presented Recomposer; as of October 2026, he works in machine perception at Google.

Now at
Google
Works
635
Citations
26,885
h-index
71
Badges

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

  1. June 1996

    An MIT doctoral thesis

    After growing up in England and earning an engineering degree at Cambridge, he completes an MIT doctorate in electrical engineering and computer science in June 1996 with the thesis Prediction-driven computational auditory scene analysis.

  2. 2000–2015

    Columbia faculty and Google

    From 2000 to 2015, he serves on Columbia's Electrical Engineering faculty and heads LabROSA, leading research in speech, environmental sound and music. In 2015, he leaves Columbia for a Research Scientist role with Google in New York.

  3. March 2017

    Announcing AudioSet

    In March 2017, he is a Research Scientist on Google's Sound Understanding Team and announces AudioSet, a human-labeled collection of over 2 million ten-second YouTube excerpts across 527 sound-event categories. The dataset provides a shared evaluation task for audio-event detection.

  4. 2018

    An IEEE Fellowship

    In January 2018, IEEE elevates him to Fellow for contributions to speech, audio and music processing.

  5. 2021

    Stronger AudioSet labels

    In 2021, he coauthors a study that collects roughly 0.1-second strong labels for 67,000 AudioSet clips. It finds that fine-tuning with weak and strong labels can improve classifier performance.

  6. 2022

    MuLan music-text embeddings

    In 2022, he coauthors MuLan, a two-tower audio-text embedding trained on 44 million music recordings and free-form text annotations. Its experiments show zero-shot music tagging and cross-modal retrieval.

  7. October 2023

    A patent for sound records

    In October 2023, Google publishes its application Enhanced computing device representation of audio, which names him as an inventor. It describes turning microphone audio into structured sound records with labels that classify non-speech environmental sounds.

  8. As of October 2026

    Google machine perception

    As of October 2026, he works in machine perception at Google, with research in audio processing, speech recognition and music information retrieval. In July 2025, he is a Research Scientist with Google DeepMind in New York and presents Recomposer, an encoder-decoder transformer for editing sound events through text descriptions and event timing.

Career

  1. Since 2017GoogleResearch Scientist
  2. 2000–2015Columbia UniversityFaculty member
  3. 1998–2001University of California, Berkeley
  4. 1997–2003International Computer Science Institute

Education

  1. 1996Massachusetts Institute of TechnologyDoctorate in Electrical Engineering and Computer Science
  2. University of CambridgeEngineering degree

From the biography and OpenAlex affiliations (the years of papers that list the organization), so some years are approximate.

  1. Google had the most people shaping AI every year from 2014 to 2025Includes Google. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  2. Google has the highest combined h-index among AI organizationsIncludes Google. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.
  3. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes Google. We estimate that 539 of the 6,766 people with a known current organization (8%) hold both a university role and a company role, including 20% of people at companies and 13% of people at universities.
  4. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Google. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Dan Ellis yet. Send us a free-licensed photo.

  1. Publications — MIT Media Lab sound.media.mit.edu. Checked 22 October 2026.
  2. Dan Ellis: Brief Bio labrosa.org. Checked 22 October 2026.
  3. thorspotify – Widening Inclusion in Music Information Retrieval wimir.wordpress.com. Checked 22 October 2026.
  4. Dan Ellis research.google. Checked 22 October 2026.
  5. Announcing AudioSet: A Dataset for Audio Event Research research.google. Checked 22 October 2026.
  6. 39 SPS Members Elevated to Fellow signalprocessingsociety.org. Checked 22 October 2026.
  7. The Benefit Of Temporally-Strong Labels In Audio Event Classification arxiv.org. Checked 22 October 2026.
  8. MuLan: A Joint Embedding of Music Audio and Natural Language research.google. Checked 22 October 2026.
  9. Enhanced computing device representation of audio patentsgazette.uspto.gov. Checked 22 October 2026.
  10. Recomposer: Event-Roll-guided Generative Audio Editing slideslive.com. Checked 22 October 2026.