What?
I propose (paid) scientific consulting services to companies willing to make the most of their data and open-source speech processing toolkits (and pyannote in particular).
I have helped companies
- optimize pretrained speaker diarization pipelines for Japanase call center recordings
- improve the accuracy of a streaming speaker diarization app
- speed up (10x) speaker diarization web APIs without compromising performance
- stay up to date with the latest speech processing technologies
Why?
- I created and maintain pyannote, the most popular open-source speaker diarization framework
- I train popular speaker diarization models
- with 1M+ monthly downloads
- 16K users or companies over the world (according to Huggingface stats)
- I design speaker diarization pipelines for specific use cases, reaching state-of-the-art performance
- 3rd place at VoxSRC 2023 (YouTube videos)
- 6th place at DISPLACE 2023 (hindi/English bilingual meetings)
- 1st place at Ego4D 2022 (egocentric videos)
- 1st place at Albayzin 2022 (TV and radio)
- (I think) I know how to share my academic knowledge and technical expertise
- my (90min long) introductory talk at JSALT 2023 summer school
- my blog post explaining my winning submissions to the above challenges
- my growing list of scientific publications
How?
We start by a free 30min intro chat to get know each other and both decide whether we want to work together.
Once a mutual agreement is reached, we collaborate either synchronously (if time difference is reasonable) or asynchronously.
- Asynchronously: we exchange via email (prefered) or messaging
- Synchronously: we meet in Zoom or Teams and I do my best to offer possible solutions to your problems (this might also need some asynchronous research on my side)
I charge by the hour and send invoices on a monthly basis.
Drop me an email at my-first-name-goes-here@niderb.fr if interested (my parents went with “herve”)