Applied RL A guide for applied Reinforcement Learning

RL introduction for LLMs, all with code that runs on a Macbook. PPO, DPO, and KL Divergence are all covered. Use your favorite IDE and LLM to go through it.

More to come!

ravinkumar.com/GenAiGuidebo...

Joining @deepmind.google.web.brid.gy as a Senior Research Engineer. Could not be more excited

Wait till you try Mariner :)

Will be at @neuripsconf.bsky.social this week giving two talks on open models and using research to make great products like @notebooklm.bsky.social. Come say hi if you're around

Yesterday I spoke for nearly 2 hours with @ravink.bsky.social (Google Labs, ex-SpaceX, Sweetgreen) in a sprawling conversation about everything generative AI and more.

Topics covered below 👇

You can watch here: www.youtube.com/live/ffS6NWq...
Vanishing Gradients podcast out next week 💫

Can I join the club? Here's some credentials

bayesiancomputationbook.com/welcome.html

You better be contributing to this trend

Building generative AI systems is NOT about choosing the right model—it’s about balancing technical depth with real-world impact. 💡

@ravink.bsky.social (Google Labs) shares insights from his journey at SpaceX, Sweetgreen, and beyond.

Full episode: vanishinggradients.fireside.fm/39

1.5 hours of my life across multiple technology fronts, from SpaceX to generative models

Evaling this model before it was released was quite fun. Glad to see it made it to the top of the people's preference

Do you also think this could be to folks using replacing "high capability" models with smaller ones?

Last year I saw people use big models by default but now I'm seeing a greater nuance as folks pick models that fit the user case. e.g. a Gemini Flash or llama 8b

From Models to Products: Bridging Research and Practice in Generative AI at Google Labs with Ravin Kumar · Luma Join Hugo Bowne-Anderson and Ravin Kumar for a live podcast recording as they discuss the rapidly changing world of generative AI, focusing on its practical…

Hugo is going to ask me hard, and fun, questions hit Thursday! Join in. lu.ma/tajz0lty?utm...

Fine Tuning Fundamentals An overview of various fine tuning methodologies

A 10 minute overview of various ways to finetune LLM models

ravinkumar.com/GenAi...

This is the first of many updates this year, with more livestreams to come!

Next session of the Intuitive AI book club is on Sunday at 8AM pst. The focus will be model safety.

All the details are in the livestream description
youtube.com/live/A-Y...

Safety is as important as model capability progression. Here's a list of references I've found helpful which come from a range of perspectives.

ravinkumar.com/GenAi...

I'm going to be using this platform more if this test post works out well! Definitely time to move to a calmer portion of the internet.

I've heard good things about namecheap

Don't search on godaddy, they're notorious for doing this

Hi all, I'll be sharing my applied "cool ways computers can do math" book club here as well. Next session is on production grade NN code and different NN architectures

community.intuitivebayes.com/t/session-3-more-complic...