Vijayasri (vij) I.

ML Scientist @ Pi School, MTech in AI, Musician, Philomath.
Interests : Multimodal Deep Learning, AI for Good, building CV and NLP powered products

[Vijayasri (vij) I. @thisisvij98]

@felixprehn AI

[Vijayasri (vij) I. @thisisvij98]

Bonus: AWS S3 also has an official PyTorch connector that allows you to load training data from your S3 bucket in a map dataset or iterable dataset style.

WebDataset: https://t.co/ryf3Wxb4nn
AWS S3 Connector for PyTorch: https://t.co/PwSmmdK9zN

[Vijayasri (vij) I. @thisisvij98]

How can you handle large datasets in PyTorch?

[Vijayasri (vij) I. @thisisvij98]

If you require a low-cost scalable setup to work with large data (especially unstructured data), AWS S3 + WebDataset can be a very good choice.

[Vijayasri (vij) I. @thisisvij98]

@FCariaggi Godsend. Really needed to see this today. I feel useful again! 😆😀

[Vijayasri (vij) I. @thisisvij98]

@avnibarman_ I'm super curious to see it !!

[Sudeepa Kolli @i_sudeepa]

@debarghya_das Wholeheartedly agree. Indian kids from middle-class backgrounds are straight-up warriors. It's puzzling to see my cousins, raised here, bothered by the smallest discomforts; I'm genuinely concerned about how they'll handle bigger challenges down the road.

[Pi School @picampusschool]

The first OpenAI Dev Day, held earlier this month, unveiled new upgrades and tools to its suite of AI products.
Pi School's Machine Learning Scientist, @thisisvij98, dives into 4 of them and explains why each enhancement is a game-changer in the evolving landscape of Artificial Intelligence.
🎥 Watch the video and join the conversation about where AI is heading next.
#MachineLearning #openaidevday #FutureOfAI

[Ankur Warikoo @warikoo]

Be prepared to look foolish infront of others, in order to learn something you don't as yet.

[Cristiano De Nobili @denocris]

The Pi School of AI programme is ready to start again!
Are interested in improving your hands-on skills in #MachineLearning, #DeepLearning and #LLMs ?
Consider to apply!

[Vijayasri (vij) I. @thisisvij98]

@sGx_tweets @PyTorch @linuxfoundation Congratulations 🎊!!

[Vijayasri (vij) I. @thisisvij98]

@_March08_ @Cmdr_Hadfield One of my favourites!!

[Vijayasri (vij) I. @thisisvij98]

There's a lot of new research on how training smaller models can also have great results, and not everyone needs a "large multi-task everything-everywhere-all-at-once model". 😀

[Vijayasri (vij) I. @thisisvij98]

With the advent of GenAI especially, we need to get comfortable with accepting the fact that our cutting-edge business advantage could come from 5-10 lines of a third-party API call. 😄

[Vijayasri (vij) I. @thisisvij98]

It's quite important to have a set of dedicated organizations "training" these LLMs at a huge scale and the rest using their checkpoints and/or training their own "smaller" models. We also need to offset the environmental impact training an LLM from scratch.

[Greg Brockman @gdb]

One nice thing about machine learning engineering is that, because the field is so new, you get to constantly invent new design patterns. Classic software engineering wisdom is still quite useful, but there's often great value in figuring out how to apply it in novel ways.

[Vijayasri (vij) I. @thisisvij98]

Many thanks @denocris !

[Cristiano De Nobili @denocris]

As a physicist, I am always bored by the non-scientific part of AI development.

However, many non-scientific aspects are of fundamental importance, such as choosing the right software licences.

This blog post by @thisisvij98 aims to explain them!
➡️: https://t.co/Vw9zRxT4JX

[Vijayasri (vij) I. @thisisvij98]

A very creative and apt comparison 👏

[Ruoshi Liu @ruoshi_liu]

Some naive thoughts about careers of AI researchers as a young uninformed PhD student  💭

An AI researcher has basically the same job as a film director. The three things you care about the most are: creativity, impact, and reputation.

(1/n)