🎙Jean-Louis Queguiner @JiliJeanlouis Twitter profile | Pikagi

Pikagi

🎙Jean-Louis Queguiner

@JiliJeanlouis

5,974

Followers

5,544

Following

1,146

Media

9,140

Statuses

Founder & CEO @gladia_io building speech AI

Paris, France

https://t.co/PFvMOEWNLQ

Joined July 2011

Don't wanna be here? Send us removal request.

Pinned Tweet

@JiliJeanlouis

🎙Jean-Louis Queguiner

4 months

Shipping a lot lately! 🚀 Been working on a series of audio intelligence features and API improvements for @gladia_io . Here are some recent releases: ▪️ Audio-to-LLM, to generate summaries, action items, and more with your own prompt ▪️ Named Entity Recognition, to identify

2

0

21

Last Seen Profiles

@BigMedicine

@Batchellor17520

@JhoanCortez7

@bokeplokalmalam

@stw_pdg

@Mariana0D

@TheeBilly

@QuinnersOnly

@UNU_IIGH

@theuwuCB

@IFPSS_DAV

@69_tiger_69

@cukienaknikmati

@TiboInShape

@BruuuhTisah

@knightpars

@zatzi

@andjelaradmilac

@XFLBrahmas

@mamadou_sir

@bvvbarendrecht

@GGMATU

@rinoaskyes

@itsFatherJoe44

@XertzyTM

@BeefEnt

@blackintheempir

@NaateR6

@becchetemmanuel

@ABer679

@blakem_22

@stw_pdg

@IrynaShev

@RepTimRyan

@shawq___0

@you_osgt24

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

MAJOR ANNOUNCEMENT Our 🎙 Audio Transcription API 🎙 is now live! Built on @OpenAI Whisper-Large-v2, 10s to transcribe a 1h file, and more accurate than everything else on the market 🔥 I’m so so proud of the team, and of the tech we created!! Why?

110

396

3K

@JiliJeanlouis

🎙Jean-Louis Queguiner

11 months

Anyone in AI and ML should read this article, TBH. It explains how to perform speaker separation, also known as diarization. This has been one of the most intriguing challenges we've faced since our inception. I love it because it illustrates that AI is largely driven by

Tweet media one

Tweet media two

Tweet media three

Tweet media four

18

165

1K

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

mind blowing ! Stable diffusion running in browser without server

Tweet card media

GitHub - mlc-ai/web-stable-diffusion: Bringing stable diffusion models to web browsers. Everything...

Bringing stable diffusion models to web browsers. Everything runs inside the browser with no server support. - GitHub - mlc-ai/web-stable-diffusion: Bringing stable diffusion models to web browser...

9

138

680

@JiliJeanlouis

🎙Jean-Louis Queguiner

10 months

Exciting news! After months of hard work at @gladia_io , we're thrilled to introduce Whisper-Zero, our latest model that solves major pain points. We asked our customers what their biggest problems were: - hallucinations (which we fixed ✅ 0 hallucinations) - language detection

Tweet media one

30

38

405

@JiliJeanlouis

🎙Jean-Louis Queguiner

10 months

Everyone talks about AI biais but I'm sure you never though about this one: At @gladia_io , we've tackled a major challenge that often goes under-discussed: language detection in the presence of strong accents. A common issue in speech recognition is misinterpretation caused by

Tweet media one

Tweet media two

Tweet media three

23

43

275

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

My neighbour has a strong GitHub contribution history

Tweet media one

2

24

260

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

It's been hell of a ride, but here we are!! Audio Intelligence API by @gladia_io is out now! ⚡️ More accurate than ever, fast, and priced to scale, with speaker diarization, word-level timestamp, and code-switching in 99 languages in the core offer.

7

34

74

@JiliJeanlouis

🎙Jean-Louis Queguiner

9 months

Most underrated Github Repo in AI + LLM JSON guided Generation:

1

29

230

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

Does anyone else remember Lotus Notes? Or am I just getting old?

Tweet media one

88

13

206

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

highly suggests everyone to watch out for what's happening there to replicate ChatGPT

Tweet card media

Open Replication of ChatGPT | Notion

Hey there 👋, nice to see you. We are working on an open replication of ChatGPT.

early-hair-c20.notion.site

3

28

187

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

just finished my API that restore photos

Tweet media one

Tweet media two

3

9

184

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

@OpenAI Go on @huggingface to try it directly

Tweet card media

High Speed Whisper Large-v2 Audio Transcription - a Hugging Face Space by Gladiaio

4

19

152

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

🎅 . @fastdotai . @math_rachel and . @jeremyphoward changed my life in so many ways that I want to repay the community 📝Comment with a project work on with . @fastdotai 🎁 I'll pick a random winner for the . @GuggerSylvain book "Deep Learning for Coders With Fastai and Pytorch"

Tweet media one

12

20

145

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

Microsoft DeepSpeed Mii is the new killer app for AI inference ... Very close to saying bye-bye to triton. Perf, Doc, Easiness, support, and time to value; all seem better.

2

12

135

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

@OpenAI We’re looking for alpha testers to keep improving on this amazing tech. Spread the word and give feedback if you can 🙏 Here’s where you sign up for the API key 👉

Gladia I Audio Transcription API

Power your product with cutting-edge AI transcription, translation, and audio intelligence add-ons using a single API. Based on enhanced Whisper ASR for companies.

32

9

132

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

@OpenAI 🚀 Inference at least 60x faster than the big Speech-to-Text players thanks to a new approach to neural network optimization - more on that here. .

Tweet media one

9

7

121

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

New Blog post ! Deepmind Alpha Tensor Explained special thanks to @ykilcher and @krypticmouse for their great content.

3

23

121

@JiliJeanlouis

🎙Jean-Louis Queguiner

8 months

I'm looking for 2 Principal Engineers: - AI Python (work on distill models, quantization, pruning, retraining/finetuning ..., triton server, cuda, ...) -> Your job is juicing as much compute power as possible from the GPUs - Low-level Architecture + Low latency optimization (JS

7

20

101

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

This @RedisLabs product is not super popular so far in data science but it will soon as we move toward inference in production keep an eye on it

Tweet card media

GitHub - RedisAI/RedisAI: A Redis module for serving tensors and executing deep learning graphs

A Redis module for serving tensors and executing deep learning graphs - RedisAI/RedisAI

4

23

97

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

Pretty cool project: use GPT3 to auto-generate commit message based on git diff.

Tweet card media

GitHub - benwr/gwipt: Automatically commit all edits to a wip branch with GPT-3 commit messages

Automatically commit all edits to a wip branch with GPT-3 commit messages - benwr/gwipt

5

11

92

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

Who’s the next Google ? something big is happening with @huggingface - over the last 1+ year 🤗 has recruited around 30 to 40 of the most amzing AI people I was following - the others were in the 20 to 30 previously hired - talents are leaving GAFAM to join 🤗 - 🤗 => 🦄 ?

@mmitchell_ai

MMitchell

3 years

👍 Article on what's been up in my career recently & @huggingface from @haydenfield ! My work there now is largely made possible by @YJernite , whose footsteps I've been privileged to follow; and increasingly in collaboration with the wonderful @SashaMTL ♥

2

15

101

6

8

91

@JiliJeanlouis

🎙Jean-Louis Queguiner

9 months

@PierreCol @fo0_ Moi je commence toujours par: “vous connaissez bloctel ?”

6

1

86

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

Number 3 arrived !!

Tweet media one

23

0

84

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

chatgpt keeps amazing me every day: I ask to edit a svg... result is perfect in the end

Tweet media one

Tweet media two

Tweet media three

Tweet media four

4

6

81

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

Best image reconstruction results I’ve ever seen

Tweet media one

Tweet media two

1

10

75

@JiliJeanlouis

🎙Jean-Louis Queguiner

10 months

Openai ChatGPT became to lazy in writing code that I moved to @huggingface chat...

7

3

70

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

We are experiencing a very high trafic and mitigating + scaling infra ! Thanks for the 🔥. we cannot believe the stats we are seeing ! We will keep you posted.

8

2

68

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

I'll miss you my friends. @OVHcloud is the best company I've ever worked for I never felt bored not a single day was the same ; If you have the chance to work there I highly suggest to give it a try. Clearly the most talented ppl I've ever met. Highest technical level + culture

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

😢😢😘🥰

Tweet media one

12

2

61

10

2

68

@JiliJeanlouis

🎙Jean-Louis Queguiner

4 years

@coqui_ai , @SpeechBrain1 and @huggingface tell your contributors that We offer 500€ gpu credit to train speech to text models twice (1K€) if it’s in another language than English (we want to make sure all language will live for fairness). Contact me on discord

3

27

65

@JiliJeanlouis

🎙Jean-Louis Queguiner

6 years

@yuvipanda Dude you’re awesome and jupyter is the best thing that happened in the last 5y for IT and Scientists keep up the good work BTw we were talking about Jupyter last week with Sylvain in my podcast. we were saying that having you is awesome 👉

0

3

64

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

😢😢😘🥰

Tweet media one

12

2

61

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

Just finalized the core of my AI side project.😊 It's a dev tool, free, open-source! Drop a 👋, I'll send you the link / details 🤪.

64

9

62

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

To transcribe audio, companies today need a huge budget and have to make compromises on quality. To solve this, we built a unique Speech-to-Text API ⚡️ 5%-10% less Word Error Rate, and up to 180x faster than the big names on the market. Read on:

Tweet media one

7

16

62

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

Just for people to kind of understand what 128K means: - 128K is around (a bit less) than 100K words - 128K is a 250 pages book - 128K is almost "The Hobbit" book (0.8x) - 128K is a bit more than 5 hours of audio conversation

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

It's really hard to believe this. just weeks after llama2 release we are already at 128K context length using YaRN scaling. I'd never imagine it would have been that fast to be honest. congrats to @NousResearch and @AiEleuther

1

3

37

3

8

60

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

🎉 Exciting News! 🎉 Last week marked a significant milestone for our team as we released another remarkable enhancement in Word Error Rate (WER) for our own Whisper's version. 🗣️✨ We're thrilled to share that we've achieved another incredible 2% reduction in WER across all

5

10

61

@JiliJeanlouis

🎙Jean-Louis Queguiner

4 years

I guess I am ready for the Wav2vec @huggingface @facebookai challenge using common voice. let's learn Speech to text for other languages than English ! let's make other langages great again (Should I retract this last sentence 🤔)

Tweet media one

5

6

55

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

OpenAI released its ChatGPT and Whisper APIs last week. Now what? My prediction: an AI price war is coming. 🧵

7

8

55

@JiliJeanlouis

🎙Jean-Louis Queguiner

4 years

Coucou... je pose ça la... bonne journée.

Tweet media one

1

14

53

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

One of my favorite pieces of software lately: - outlines: - llama index:

Tweet card media

GitHub - dottxt-ai/outlines: Structured Text Generation

Structured Text Generation. Contribute to dottxt-ai/outlines development by creating an account on GitHub.

1

11

52

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

@OpenAI ✅ Word Error Rate (WER) between 1% - 3 % on human-readable transcripts using #OpenAIWhisper Large v2, that is ≈ 5x better than other major providers.

3

3

52

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

Just saying

Tweet media one

6

7

50

@JiliJeanlouis

🎙Jean-Louis Queguiner

4 years

Here we go ! @OVHcloud #AI training platform is out !!! 1.75€/h for a V100s 32GiB/GPU | 16.4 TFlops 14vCores 25GB RAM Auto-tiering on object storage no need for object storage gateway Only 1 cli to rule them all

Tweet media one

Tweet media two

3

27

49

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

Since we launched our #alpha yesterday, some have asked: why #audio now? Because the market is broken. Here's what I mean 🧵

5

6

52

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

I'm looking for a freelance web designer anyone to recommend me ? It's for an awesome AI project. c.c. @BastiUi @benjamincode

40

3

36

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

@pbeyssac @PartiPirate Sinon solution française : @gladia_io full hébergée Europe ( @OVHcloud ). Bcp plus précis et rapide (whisper hallucine bcp on a tout une mécanique pour gérer ça).plus précis encore si vous envoyez du wav 16KHz. Ca fait la séparation des speakers, les exports sous-titres en srt/vtt

3

4

51

@JiliJeanlouis

🎙Jean-Louis Queguiner

5 years

@streamlit is awesome ! check out our @OVHcloud Image Processing API (quick and dirty lab). In 2 clicks you can : #ImageRecognition #Colorization (based on @citnaj 's work) #Segmentation #NeuralStyle #selfie2Anime + many more! Have fun here =>

8

13

48

@JiliJeanlouis

🎙Jean-Louis Queguiner

5 years

Thanks to @Deezer research we were able to provide #spleeterAsAService the Sound source separation API that will split Vocals / Piano / Bass / Drums from a soundtrack. Enjoy on @OVHcloud

Tweet media one

3

20

47

@JiliJeanlouis

🎙Jean-Louis Queguiner

5 years

Je cherche un stagiaire Machine Learning pour travailler sur la conception et l'implémentation d'une stratégie de détection de la dérive conceptuelle lors de l'inférence d'un modèle.

4

41

47

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

We just integrated a v0 of diarization (speaker separation) on top of whisper large v2 api transcription from @gladia_io We still have a few glitches to fix yet pretty cool. Also for people that were looking for plain text output this is the place !

Tweet card media

High Speed Whisper Large-v2 Audio Transcription - a Hugging Face Space by Gladiaio

3

6

46

@JiliJeanlouis

🎙Jean-Louis Queguiner

11 months

Live transcription by @gladia_io is out in general availability 🔥 I recommend you check it out: its accuracy in real-life use cases is impressive, including w/ timestamps and code-switching. As a feature that's not supported at all with vanilla Whisper, it's a great milestone

Tweet media one

4

5

45

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

Hi 👋 if you are interested in: 🐍Python 👨‍💻AI APIs for web devs 🤖AI/Machine Learning 💾Data/Big Data 🧑‍🔬Quantum Computing Follow me. ✔ I use Twitter to share a lot of content that you won't want to miss. 🎉

1

11

44

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

@AtaxyaNetwork 1/ peu de gens dans le monde du réseau sont sur Twitter et expose ce qu’il font 2/ rare sont les photos de DC 3/ tu as partie de 0 en expliquant et grandissant avec la communauté donc tout le monde peut se projeter c’est un peu comme un tuto géant ;-) 4/ tu râles pas

1

0

45

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

What a week! 🎉 On behalf of myself and the @gladia_io team, THANK YOU to all of you who've been giving us feedback since the🎙️ alpha launch🎙️! We're in awe of how passionate and supportive the community here is and look forward to building a perfect final product for you all 💙

2

4

44

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

N’oubliez pas de couper votre wifi

@TripInChina

Sharing Travel

2 years

Night in Wuhan, Jianghan District.😍😍😍

266

1K

6K

4

4

42

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

My last day @OVHcloud = My last Blog Post @OVHcloud It talks about #GPU #CPU #VM #OpenStack #VMWare #NVIDIA + a bit of security around CPU Please share your thoughts in comments !

Tweet card media

How to virtualize PCI-e GPU in the cloud ? - OVHcloud Blog

A quick introduction to virtualization The purpose of virtualization is to isolated the user software environment from the hardware environment. The orchestration between these virtual environments...

blog.ovhcloud.com

2

8

43

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

You asked for it, we delivered! 🎤 Speech-to-Text AI #translation , as first among a series of new features we're working on at @gladia_io . Same core API, more capabilities. Register for the #alpha now if you still haven't!

@gladia_io

Gladia.io

2 years

New feature alert!🎤 You can now use our core audio transcription alpha to #translate speech from audio! Just pick an output language and our #API will do the rest ✨ Available in 99 languages 🌍 #ai #stt #transcription #whisper

3

5

23

4

7

42

@JiliJeanlouis

🎙Jean-Louis Queguiner

5 years

Hi everyone here is the new @OVH #AI #Marketplace ! check this out and let us know your feedbacks #AI #IA #MachineLearning

2

29

42

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

With @tom_wilkie CTO of @grafana !! If you have told me that 1 day I’ll meet Grafana’s CTO…. Who happened to write Loki.

Tweet media one

1

1

43

@JiliJeanlouis

🎙Jean-Louis Queguiner

11 months

Following the fantastic event by @motierventures , it is super exciting to see @gladia_io listed alongside @dust4ai @MistralAI @Scenario_gg @finegrain_ai as part of the booming French AI ecosystem. Doing this alongside you all is a pleasure and privilege.

Tweet card media

There’s something going on with AI startups in France | TechCrunch

Artificial intelligence, just like in the U.S., has quickly become a buzzy vertical within the French tech industry.

3

5

43

@JiliJeanlouis

🎙Jean-Louis Queguiner

5 years

@davidperrotin Pour ceux qui débarquent comme moi :

2

14

41

@JiliJeanlouis

🎙Jean-Louis Queguiner

11 months

Wrapping up the week with the release of speaker diarization v2 by @gladia_io !📢 Since we launched in alpha, it's been the most popular feature among the users - so we committed to providing the best quality there is on the market. The new version has been rebuilt from the

Tweet media one

2

9

38

@JiliJeanlouis

🎙Jean-Louis Queguiner

6 years

Montage de l’épisode 7 terminé : il est trop bien. Le thème c’est #kubernetes pour les nuls. Merci à @LostInBrittany @0xd33d33 et @StevenLeRoux . J’ai passé une heure géniale ! on se retrouve mercredi prochain 18:15 en attendant bonne écoute

Tweet media one

3

16

37

@JiliJeanlouis

🎙Jean-Louis Queguiner

4 years

Looking for the @fastdotai of quantum computing : @pennylaneai is the way to go. It’s clear with all you need at one place : 1/ a simple wrapper with associates boiler plate 2/ a strong community 3/ code oriented first rather than fully theorical 4/ very clear educational content

1

10

39

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

OVHCloud AI Training : le géant français du Cloud lance son offre IA et Machine Learning

Tweet card media

OVHcloud AI Training : le géant français du Cloud lance son offre IA et Machine Learning

OVHCloud, le géant français du Cloud Computing, lance l'offre AI Training pour l'entraînement d'intelligence artificielle et la Data Science.

www.lebigdata.fr

4

18

38

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

Despite the expansion of @huggingface with transformers I still believe that @spacy_io is one of the underrated ML/AI treasures.

2

1

35

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

It's really hard to believe this. just weeks after llama2 release we are already at 128K context length using YaRN scaling. I'd never imagine it would have been that fast to be honest. congrats to @NousResearch and @AiEleuther

Friendly reminder, the guys actually designing high performance Data Structures are yeeting their alpha for free https://t.co/MMJLltgvVS Meanwhile there is a parasitic class that preys on the...

1

3

37

@JiliJeanlouis

🎙Jean-Louis Queguiner

6 years

And here we are !! #MachineLearning platform at @OVH ! This is doing all the feature preparation work for you as well as the #autoML and the model serving through #API in just few clicks... Instill can believe we did it.

2

29

36

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

Very good lib, I think most people misunderstood the point of the license. Most people complaining about the new license are very likely not affected. I also read comment “so basically you want free work from the community”. While I understand that people are p***ed. I think

@jeffboudier

Jeff Boudier 🤗

1 year

Today is a huge milestone for one of our latest libraries: Text Generation Inference - we released v1.0 and under a new license: HFOIL 1.0 This 🧵 explains what this new license means, and why the change!

17

26

123

1

6

36

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

gladia_io: @OpenAI ✅ And it’s top-quality! We achieved a Word Error Rate (WER) between 1%-3% on human-readable transcripts, based on real-life use cases. That’s ≈ 5x better than other major providers.

Tweet media one

1

4

36

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

Our seed round and product launch are featured in @TechCrunch ! And according to @romaindillet , @gladia_io is the best transcription API he's ever used...We're humbled and thrilled. Off we go team!!

Tweet card media

Gladia turns any audio into text in near real time | TechCrunch

Gladia is a French AI startup that wants to change how companies interact with audio data. The company develops an audio transcription API.

7

9

35

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

If you are building @tiangolo fastapis you shoudl consider adding prometheus instrumentation to monitor : api latency, nb of call etc... c.c. @philipvollet

Tweet card media

GitHub - trallnag/prometheus-fastapi-instrumentator: Instrument your FastAPI with Prometheus...

Instrument your FastAPI with Prometheus metrics. Contribute to trallnag/prometheus-fastapi-instrumentator development by creating an account on GitHub.

3

6

34

@JiliJeanlouis

🎙Jean-Louis Queguiner

4 years

Tech porn

@bob_burrough

Bob Burrough

4 years

What’s the best way to apply thermal paste? 👇

87

1K

4K

1

14

33

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

@MathisHammel En terme de management : le mec cherche juste à écrémé pour le moins cher possible. Ça te soûl tu vas te barrer, ceux qui sont encore là après vont se faire auditor par les mecs de tesla: ça te soûl tu vas te barrer. A chaque fois il paie pas le package de sortie. Donc malin…

6

1

33

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

Stylegan / CLIP (STYLECLIP) be like

0

4

34

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

gladia_io: @OpenAI Best part? We released an alpha version for free! You can register on Gladia at . We count on your feedback to make it perfect! And pls spread the word if you like it 🙏 We want as many of you as possible to …

Tweet media one

1

8

33

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

Voir des meme comme ça me fait du bien ça me redonne foi. On va peut être arriver à déconstruire les a priori sur le sujet et aller de l’avant.

@Gen_Atomic

Generation Atomic

3 years

#MemeMonday #NuclearEnergy #NuclearPower #CleanEnergy #meme #memes #bestmemes

Tweet media one

57

1K

4K

5

13

33

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

YouTube URL transcription ☑️ #AI #stt #gladia

@gladia_io

Gladia.io

2 years

Let's face it: automatic captions on #YouTube are not great. But we've got something to help you with that. Use @gladia_io API to turn your videos into high-quality #transcriptions in seconds and improve your viewers' experience. Just drop a URL, our AI will do the rest ✨

6

3

30

3

2

30

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

📢 𝐉𝐮𝐬𝐭 𝐑𝐞𝐚𝐝: "Extrapolating Large Language Models to Non-English by Aligning Languages". A deep dive into making LLMs more multilingual. 🌍🔍 Paper: GitHub Repo (though it's empty 😞): 🧠 𝐛𝐚𝐜𝐤𝐠𝐫𝐨𝐮𝐧𝐝: LLMs, like

Tweet media one

Tweet media two

Tweet media three

Tweet media four

0

1

31

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

We're currently working on audio transcription for emergency services. And I can tell you one thing... These people have a tough job. Listening to despairing calls all day long is so unsettling. People on the other end of the line panicking while the fire is approaching,

3

3

31

@JiliJeanlouis

🎙Jean-Louis Queguiner

5 years

We love @mybinderteam and @ProjectJupyter so much that we decided to be part of the contribution! Thanks @mybinderteam for letting us be part of the amazing story you are writing !!

@BastienVdbt

Bastien Vé @baaastien.bsky.social

5 years

Very proud to see @OVH helping the @ProjectJupyter community ! We just joined the @mybinderteam federation and we share the same values. Read more here

1

17

36

0

10

31

@JiliJeanlouis

🎙Jean-Louis Queguiner

4 years

With these 3 plots from @weights_biases we can clearly see the autoML searching for the right learning rate for 6 epochs and once it's found starting to learn for real. @weights_biases is the gitHub of DataScience. Awesome product.

Tweet media one

2

9

30

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

I'm looking for a #python Developer (junior accepted).

9

23

27

@JiliJeanlouis

🎙Jean-Louis Queguiner

6 years

Les sanctions de la CNIL en infographie

Tweet media one

2

17

30

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

Big tech/product update @gladia_io 🔥 Whisper theory vs. reality Theory: 🤔 "Whisper can do word level timestamp." Reality: 🎯 Precision is terrible 😵 When dealing with word-level timestamps within Whisper, it will drop some words/parts of speech ⏱️ Compute time can be up to

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

New Release on @gladia_io :

Tweet media one

1

3

16

5

5

31

@JiliJeanlouis

🎙Jean-Louis Queguiner

7 months

We’re working on something very exciting at @gladia_io : audio intelligence add-ons, to build robust apps with on top of speech-to-text. Summarization is out in alpha now, and a lot more is coming soon, like chapterization, NER, PII redaction, etc. You can request early

Tweet media one

3

3

31

@JiliJeanlouis

🎙Jean-Louis Queguiner

4 years

0

18

31

@JiliJeanlouis

🎙Jean-Louis Queguiner

10 months

Loved being part of ai-PULSE conference by @Scaleway today ! 🚀 Impressive level of technical expertise and granularity of insights exchanged by all speakers. Pleasure sharing the panel with @jwittich @gcambefort and Maryan from @Powder_gg . The French AI tech is up and running !!

Tweet media one

1

6

31

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

Had to refactorized a bit the backend CI/CD module last week. Using @OVHcloud open source Continuous Delivery System I am preparing a blog post + video serie explaining how to setup it up and how it’s saving me hours of compute and k$ of GPU money.

Tweet media one

1

9

30

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

Je cherche un full stack avec bonne connaissances front end pour bosser avec moi. Des sujets de fou. Petite team Paris / rennes/Brest/Lille/Toulouse/bordeaux/Lyon.

6

24

29

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

Our first community-led roadmap is out! Thanks again for all of your product suggestions, building with community is lots of fun 💙 We're on track to make @gladia_io API even more robust in the following weeks.

@gladia_io

Gladia.io

2 years

New month, new features 🚀 We're happy to present our📍March '23 roadmap📍feat. top 3 most anticipated features requested by the alpha community! First release coming later this week. #gladia #ai #stt #api

Tweet media one

1

6

36

2

2

29

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

so far so good keep pushing

Tweet media one

4

0

28

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

Emotion recognition is out!! The most humane source of insights into your calls, meetings, and media content. Here's what @gladia_io detects: happiness, sadness, anger, fear, surprise, and disgust. Play with it, and let me know!

1

3

29

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

Adobe is really silent : people from @Adobe and associates university are not shouting out everywhere saying their results are the best But… they are… honestly when it comes to image to image AI Adobe is Waaaayyyy ahead of anyone else! GG @Adobe I really admire your work

@jbhuang0604

Jia-Bin Huang

3 years

Hey! Got 40 seconds? ⏱️ Learn how we achieve photorealistic reposing and virtual try-on in the upcoming SIGGRAPH Asia paper *Pose with Style*. 🤩 Paper: Web: Brought to you by the amazing @BadourAlBahar !

4

68

316

2

5

29

@JiliJeanlouis

🎙Jean-Louis Queguiner

2 years

@KashPrime @OpenAI this is something we are still iterating on but our mission is to make AI accessible and we found that the Speech to Text market was broken talking to users: For instance, the pattern is always the same when it comes to call-center STT.

Tweet media one

2

1

27

@JiliJeanlouis

🎙Jean-Louis Queguiner

6 years

Mmmm I’ve never seen it that way ! ⁦ @Defensivesec ⁩ ⁦ @haveibeenpwned ⁩

Tweet media one

4

14

26

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

Our longer-term vision is to move from 2D to 3D data. Audio can be pretty flat, and the idea is to augment it with intelligence. This is what our new brand is about. Hope you enjoy it as much as we do:

0

5

28

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

16k and 32k released the same day. Open source is on fire!

@togethercompute

Together AI

@togethercompute

1 year

We just released LLaMA-2-7B-32K, a 32K context model that can be fine-tuned for tasks like doc understanding, summarization & QA! Built with Position Interpolation & our data recipe/optimizations, run inference & fine-tune with up to 3x speedup. Thread👇

12

168

723

1

4

28

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

Better understanding of @OVHcloud @MongoDB offer vs others

Tweet media one

0

7

28

@JiliJeanlouis

🎙Jean-Louis Queguiner

1 year

Ever wondered how #prompt injection in speech recognition work? Here comes my deep dive 🏗️ Let's start with an overview of audio processing in Automatic Speech Recognition (ASR)

Tweet media one

2

2

27

@JiliJeanlouis

🎙Jean-Louis Queguiner

3 years

😀Person to Follow of the week : @philipvollet because : - He has done so 💅 for the AI/ML open source community - He has always been there to help 🙏 Follow @philipvollet to keep it up with @spacy_io , NLP, open source for AI/ML in general.

0

4

25