Karan Goel
CEO and founder of Cartesia; co-author of the S4 and SaShiMi state space model papers
Karan Goel is the CEO and a founder of Cartesia, whose products include the Sonic text-to-speech and Ink speech-to-text models. Goel's PhD work at Stanford, advised by Christopher Ré, included co-authoring the S4 paper on structured state spaces and first-authoring SaShiMi, which applied state space models to raw audio generation.1345
Career
Goel studied for a computer science PhD at Stanford, advised by Christopher Ré and affiliated with the Stanford AI Lab and the Center for Research on Foundation Models. Goel's site notes early involvement in the Stanford Mercury project, which became the Center for Research on Foundation Models.3
Cartesia's company page lists Goel as CEO and founder, alongside co-founders Albert Gu, Arjun Desai and Brandon Yang, and says the founding team met as PhD students at the Stanford AI Lab, where they worked on state space models. Cartesia's products include the Sonic text-to-speech and Ink speech-to-text models.1
Research
Goel's PhD research covered model auditing and robustness (Robustness Gym, Mandoline and Model Patching), interactive data systems for machine learning (Meerkat), and sequence modelling with state space models. Goel co-authored S4 (2021) with Albert Gu and Christopher Ré, was first author of SaShiMi (2022) on audio generation, and co-authored S4ND (2022) on images and video.34567
Goel also helped build the Stanford MLSys Seminar series and organized Data-Centric AI workshops, according to Goel's site. Cartesia's research page lists these papers.32
Work
It's Raw! Audio Generation with State-Space Models (SaShiMi) (external site: arxiv.org)Published research
First author of the 2022 paper introducing SaShiMi, an architecture that applies state space models to raw audio waveform generation.53
Efficiently Modeling Long Sequences with Structured State Spaces (S4) (external site: arxiv.org)Published research
Co-author, with Albert Gu and Christopher Ré, of the 2021 S4 paper on structured state space sequence models.42
S4ND: Modeling Images and Videos as Multidimensional Signals Using State Spaces (external site: arxiv.org)Published research
Co-author of the 2022 paper extending state space models to images and video.62
Career timeline
Aug 2020
Posted "Model Patching, Closing the Subgroup Performance Gap with Data Augmentation" as first author.7
Oct 2021
Co-authored the S4 paper on efficiently modeling long sequences with structured state spaces.4
Feb 2022
Posted the SaShiMi paper on audio generation with state space models as first author.5
Oct 2022
Co-authored S4ND, on modeling images and videos as multidimensional signals with state spaces.6
Other documented roles
Roles the sources document but that are not confirmed as current.
- Computer Science PhD student in the Stanford AI Lab, advised by Christopher Ré, Stanford University31
Related people
More sourced profiles at the same organization, in the same field, or from the same school.
- Albert GuAlso at Cartesia
- Alec RadfordLanguage models
- Abigail SeeAlso at Stanford University
- Arjun DesaiAlso at Cartesia
- Dan JurafskyAlso at Stanford University
- Abubakar AbidAlso from Stanford
- Brandon YangAlso at Cartesia
- Dong YuSpeech and audio
- Andrej KarpathyAlso from Stanford
- Andrew NgAlso at Stanford University
- Edouard GraveLanguage models
- Awni HannunAlso from Stanford
Is this you?
Profiles cover public professional work only, and every line cites a source. If something is out of date, missing or wrong, send a correction with a link to an official page and we will update it. Suggest a correction
Add this badge to your website or GitHub profile README. It links back to this page. It is not an endorsement or a rating.
Markdown (for a README)
[](https://unitedstatesofamericasuperintelligence.com/people/karan-goel/)
HTML
<a href="https://unitedstatesofamericasuperintelligence.com/people/karan-goel/"><img src="https://unitedstatesofamericasuperintelligence.com/badges/profile.svg" alt="Karan Goel — Profile on USASI" height="20"></a>
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.