Neha Pawar

NLP Enthusiast, Applied Research, Datascientist @DATEV

[Google AI @GoogleAI]

Stop by the #ICLR2024 Google booth today at 3:15 PM to hear @rach_it_ and @nitish_gup talk about Composition to Augment Language Models (CALM), which introduce cross-attention between LLMs to compose their representations and enable new capabilities.

[Neha Pawar @traintestrepeat]

@zehavoc @mayhewsw Except that express in some parts of the world mean <90km/hr . Highlights the need to have an uniform understanding where the LLMs begin?

[Neha Pawar @traintestrepeat]

#DATEVLab remembers it's friends who helped us along the way. YES to #innovation, YES to the #future.

[Neha Pawar @traintestrepeat]

@IamSubhayu Subhayu, it breaks my heart to read the trouble you are going through. As a fellow NRI, I can relate to what you are feeling. I am hoping the tough times are over soon.

[Neha Pawar @traintestrepeat]

@ChristianKlose3 The lottery ticket hypothesis is impressive. It was also my Tweet debut ;)

[François Chollet @fchollet]

There are two kinds of software developers: those who already use deep learning, and those who will use deep learning next.

[Neha Pawar @traintestrepeat]

#DATEV has more than 170 products in the tax, accounting an the legal domain. Imagine the immense potential AI can deliver. In the #AILab #Readinggroup we #DATEVianers discuss and debate the published research.

[DATEV @DATEV]

#TECHBLOG: Die Reading Group unseres #DATEV #AI Lab feiert ihr Einjähriges.
Einen Blick auf die Themen der Zukunft wirft unsere Kollegin Neha Pawar (@traintestrepeat) im #TechBlog. https://t.co/4zf30UMfXq

[Ferenc Huszár @fhuszar]

Perserverance, you're on mute.

[tkasasagi 🐻 @tkasasagi]

How about no.

[Neha Pawar @traintestrepeat]

If this hasn't happened yet then one has'nt experimented enough! Next step: hoping against all odds that the bug is not real !!

[Mark Riedl @mark_riedl]

TFW you evaluate your AI system and it performs 1000% better than the best available baseline.

[Neha Pawar @traintestrepeat]

@DhirajS65167098 Famous and the one that works may not the same. In NLP, there is also no one shoe that fits all. I follow the following steps: understand the usecase, read on the concepts that work for the problem, experiment and build an intuition for the various problems. I hope this helps

[Neha Pawar @traintestrepeat]

@DhirajS65167098 Famous and the one that works for may not the same. In NLP, there is also no one shoe that fits all. I follow the following steps: understand the usecase, read on the concepts that work for the problem, experiment and build an intuition for the various problems. I hope this helps

[Neha Pawar @traintestrepeat]

@DhirajS65167098 When you hit a wall or want a solution more specific to your usecase you will feel the need to learn tensorflow or pytorch. n/n

[Neha Pawar @traintestrepeat]

@DhirajS65167098 If you're at early stages, it can be sometimes frustrating. But, persevere and I promise it gets better. The library you use will heavily depend on your usecase. I would advice you to start off with a high level library for rapid prototyping. Spacy is a good choice. 2/n

[Neha Pawar @traintestrepeat]

@elonmusk suggests Signal (the app) as an alternative for whatsapp and stocks spike up for an unrelated company called Signal > 0.6 to 7$ right now. Pure fun for anyone trying to perfect the stock prediction algorithm :D

[K10✨ @Kristennetten]

@elonmusk @elonmusk this poor company has just been launched from .60 a share to 3.60 + due to people thinking you meant Signal the Company $SIGL & not the social app. 😅

[Neha Pawar @traintestrepeat]

Received these amazing stickers. Thanks! @spacy_io @explosion_ai

[Neha Pawar @traintestrepeat]

People matter!

[@timnitGebru (@dair-community.social/bsky.social) @timnitGebru]

I cannot believe that this has actually happened. And the letter is so thorough. Thank you!!!

[(((ل()(ل() 'yoav))))👾 @yoavgo]

is this... a surprising finding?

https://t.co/PduyPdURsB