Antoine Sinton

@Artisans_Cartos Trouve !

Ces 2 bâtiments font partis des plus grands de France. Plus de 230 000 m² chacun ! D'après vous, qu'est-ce que c'est ?

Where am I going to get my fix of rainbow poops now??? Working for @zenly was a super intense experience, it'll stick with me forever. They're all amazing at what they do, so if you have the chance to hire one don't hesitate for a second.💔🍦

@hugodouchet c'est expliqué dans cet article rtl.fr/culture/arts-s…

Pourquoi dit-on Onze et pas Dix et un, comme on dit Vingt et un ? Si quelqu’un a la réponse ça m’intéresse, c’est pour un ami.

"🚀Fundraising Alert🚀@ContinuityTech raised a €5m seed investment led by us & @Bpifrance🙌 🤖Continuity’s #SaaS platform uses #OpenData & #AI to empower P&C #SME insurers & help underwriters deliver smarter and faster! frenchweb.fr/seed-continuit…

@julsimon @huggingface Yes exactly, especially when using spot instances. The pressure on GPU spots right now is huge whereas on CPU instances it's a lot less. There is a tradeoff here where spending a little more time training on CPU actually means getting the output more quickly.

@poudro @huggingface Yes, always good to have plenty of options.

The rise of transfer learning is changing the game, and distributed training on Intel CPUs is an attractive, cost-effective option. Learn more in this new @huggingface blog post huggingface.co/blog/accelerat… #MachineLearning #DeepLearning #MLOps #MLEngineering

@julsimon @huggingface interesting, thanks

@poudro @huggingface Yes, I ran some tests using the latest DLAMI (PyTorch 1.10, not 1.9). p3.8xlarge: train_runtime = 0:00:40.93, g5.24xlarge: train_runtime = 0:01:19.22. Would be interesting to try smaller c6i sizes as well. Pretty sure 8xl and maybe even 4xl would do a good job.

@PluClaire @zenlyapp Recently got one, it's like nothing I've sat on before

@PluClaire @zenlyapp Herman Miller Aeron!!!

Ok Twitter, I need your help 🙏 We’re looking for the ✨ very best office chair in the world✨ for the @zenlyapp team. Whoever recommends the one we finally go for, will receive a special thank you from us. 👇👇

Hello twitter 👋! We're looking for computer vision engineer/scientist for a mission on super-resolution. Let me know if you are interested!

@MarcdeFalco sinon il y a toujours gnuplot 😬 bon ok je sors ➡🚪

@DFintelligence Totally agree! Even efficiently following SOTA is not possible across all domains.

What are you thougts ? #datascience #IA #MachineLearning #artificialintelligence #intelligenceartificielle 6/6

Why a Data Scientist should not be an expert in all Data Science fields ?....... A THREAD WITH GIF 1/6

We used a transformer model to increase the performance of our postal address parsing pipeline link.medium.com/ij1TGvAe4ib

@ClementDelangue I would say @RecitalAI and, also, less of a startup nowadays but they are doing interesting stuff: @Deezer cf @researchdeezer Also, us, Continuity (no twitter presence), we're super small still, but hopeful to build some cool ML in the insurtech space

What are the best machine learning startups in Paris these days?

super cute @zenlyapp love it! ❤️💉

@MarcdeFalco Je me souviens avoir dévoré le Gleick

Est-ce qu'il existe un captcha "responsable" ? Càd un captcha où les annotations générées sont d'utilité publique (climat, éducation, recherche publique, etc...)

@EmilyPython @MarcdeFalco Je pense qu'il faudrait interdire π 🙊

@MarcdeFalco Pas d'inquietude, il y aura rapidement une startup pour deployer une solution de MechanicalTurkSort dans le plus pur respect de la GDPR et des chartes de la CNIL ! On est sauf !

@MarcdeFalco Mais du coup, ils veulent encadrer l'usage du quicksort par exemple ? C'est bizarre ce qu'il dit le monsieur :D