Vasudev Sharma

MLE | MScAC @UofT | Tweets about Visualization, AI and ML 🤖 and productivity. Views are solely mine.

[Vasudev Sharma @vasudev_sharma_]

@cursor_ai @grok are technical details of cursor 2 out?

[Vasudev Sharma @vasudev_sharma_]

@danifesto Memory

[Vasudev Sharma @vasudev_sharma_]

A side project I have been working on recently: a ML pipeline that automates developing ML models with Vertex AI for both batch and online inference

https://t.co/qA3WnqFpjo

[Recursion @RecursionPharma]

Next stop: LONDON! 🇬🇧

We’re expanding our operations to Europe, at the epicenter of the rapidly growing #TechBio sector in #London.

Learn more at https://t.co/qhw07NKdUC and check out our new open positions at https://t.co/f7aGJTXkOe $RXRX #biotech #ai #ml

[Vasudev Sharma @vasudev_sharma_]

Standards are like toothbrushes. Everyone was to have one, but no one wants to use someone else's.

[Valence Labs @valence_ai]

Introducing LOWE: an LLM-Orchestrated Workflow Engine for executing complex drug discovery workflows using natural language.

The next evolution of the @RecursionPharma OS, LOWE has access to proprietary data, integrates with Recursion’s labs, and more 🧵

https://t.co/mCJDe9z9xZ

[Chris Gibson @RecursionChris]

Biology transitioned from Art —> Science in the last century.

In this century (and really the next decade) it will transition again from Science —> Engineering.

As @nvidia’s Jensen Huang shares, this shift will underlie extraordinary opportunity.

I am excited to be working at the leading edge of this transition with companies like $NVDA as there are few better places to build with more upside and potential for positive impact…

[Vasudev Sharma @vasudev_sharma_]

*Hiring alert*

Our team (ML Infra) at @RecursionPharma is hiring for Sr. Machine Learning Engineer role. This role is open both in US and Canada.

If you're interested in solving challenging problems in the space of AI/ ML and biology, apply here ⬇️

https://t.co/UWJfFBs81b

[Vasudev Sharma @vasudev_sharma_]

Something interesting is cooking within our team at Recursion 👀

[Chris Gibson @RecursionChris]

In a Large Language Model like @OpenAI's ChatGPT , a model has been trained to predict the best next token (e.g., word) to add iteratively in a written response. As these models get trained across more and more data, they start to learn foundational elements of text which, when fine-tuned, can sometimes feel almost sentient!

At @RecursionPharma, our focus isn’t text, but biology and chemistry! Like the LLMs you see in the news, we are similarly training AI models that can learn something foundational about biology through a simple image-based task. We’ve trained these models across more than a billion proprietary biological images of human cells we’ve generated in our own wet-labs and then we ask the model to predict the missing parts of an image (reconstruction) when we mask 75% of it (masked input).

At the scale of our data, you can see our models have gotten pretty good at this task! But what’s the utility of predicting masked images of human cells by itself? Like an LLM, these models learn foundational things about biology that help produce emergent capabilities (e.g. they are better at detecting subtle cellular phenotypes, identifying relationships between biological perturbations, etc) that are ultimately helping us make better predictions to drive what we hope will be better potential medicines!

We will have a lot more to share soon… stay tuned ;)
#ShowMeTheModel

[Chris Gibson @RecursionChris]

In a Large Language Model like @OpenAI's ChatGPT , a model has been trained to predict the best next token (e.g., word) to add iteratively in a written response. As these models get trained across more and more data, they start to learn foundational elements of text which, when fine-tuned, can sometimes feel almost sentient!

At @RecursionPharma, our focus isn’t text, but biology and chemistry! Like the LLMs you see in the news, we are similarly training AI models that can learn something foundational about biology through a simple image-based task. We’ve trained these models across more than a billion proprietary biological images of human cells we’ve generated in our own wet-labs and then we ask the model to predict the missing parts of an image (reconstruction) when we mask 75% of it (masked input).

At the scale of our data, you can see our models have gotten pretty good at this task! But what’s the utility of predicting masked images of human cells by itself? Like an LLM, these models learn foundational things about biology that help produce emergent capabilities (e.g. they are better at detecting subtle cellular phenotypes, identifying relationships between biological perturbations, etc) that are ultimately helping us make better predictions to drive what we hope will be better potential medicines!

We will have a lot more to share soon… stay tuned ;)
#ShowMeTheModel

[Ben Mabey @bmabey]

Big news! This collaboration with @NVIDIA marks a significant milestone on @RecursionPharma's journey toward revolutionizing drug discovery through AI. I'm thrilled to be working with them on pushing what is possible with AI in drug discovery.

[Vasudev Sharma @vasudev_sharma_]

@ahmadphosseini @UofT Congrats 🎉

[Vasudev Sharma @vasudev_sharma_]

@GoogleColab That's huge!!

[Master of Science in Applied Computing (MScAC) @UofTMScAC]

Congratulations to our MScAC students as they celebrate their graduation this week! We spoke with ten of our new grads who've shared with us their student experience, passions and motivations for pursuing the @UofTMScAC program.

Read more: https://t.co/D4JAW8jNRW

#UofTGrad23

[Vasudev Sharma @vasudev_sharma_]

@csandyash Up and onwards 🔥

[Vasudev Sharma @vasudev_sharma_]

@EdkeParinita @RecursionPharma Thanks :)

[Vasudev Sharma @vasudev_sharma_]

Excited to share that I have started working as ML engineer @RecursionPharma with the aim of decoding biology with AI and ML. Exciting times ahead. 😄

[Vasudev Sharma @vasudev_sharma_]

@rahulgk Congrats @rahulgk 🎉

[Vasudev Sharma @vasudev_sharma_]

Keeping abreast of the research in AI space is impossible. Goal should be to ignore the distractions than keeping up with the current pace.

[Gowthami @gowthami_s]

This is exactly how I've been feeling for the last few months. It is becoming almost impossible to keep up with the current pace of research.

#phdlife #machinelearning

[Vasudev Sharma @vasudev_sharma_]

The bird is back again 😄

COURTESY: ELON MUSK

[Vasudev Sharma @vasudev_sharma_]

Is my twitter hacked?