Arvind Rajaraman

LLMs x systems @databricks • Opinions my own

[Arvind Rajaraman @arvindr02]

@OpenAI congrats @SeyoneC!

[Arvind Rajaraman @arvindr02]

@pgasawa congrats on the launch!

[Arvind Rajaraman @arvindr02]

@pgasawa cool work!! test-time training is surprisingly underexplored

[Arvind Rajaraman @arvindr02]

@tonyzzhao @sundayrobotics Congrats Tony!!

[Arvind Rajaraman @arvindr02]

@pgasawa Training a ParthGPT rn

[Parth Asawa @pgasawa]

Training our advisors was too hard, so we tried to train black-box models like GPT-5 instead. Check out our work: Advisor Models, a training framework that adapts frontier models behind an API to your specific environment, users, or tasks using a smaller, advisor model (1/n)!

[Arvind Rajaraman @arvindr02]

@brandontrabucco congrats!

[Arvind Rajaraman @arvindr02]

@oetanorev my first thought was which Jax library is that

[Arvind Rajaraman @arvindr02]

@marsupialtail_2 🙋‍♂️

[Rak Garg @rak_garg]

Sitting next to a Databricks engineer on Caltrain who's crushed three Celsius cans, is typing with the fury of a thousand Devins, and has not made eye contact once. This is what dreams are made of

[Matei Zaharia @matei_zaharia]

Excited to launch Agent Bricks, a new way to build auto-optimized agents on your tasks. Agent Bricks uniquely takes a *declarative* approach to agent development: you tell us what you want, and we auto-generate evals and optimize the agent.

https://t.co/EVqwq583cF

[Arvind Rajaraman @arvindr02]

@zeeshanp_ @xai congrats!

[Arvind Rajaraman @arvindr02]

@ekzhang1 @_shreya_s thanks for hosting!!

[Arvind Rajaraman @arvindr02]

@CadeGordonML @Anthropic LFG so well deserved

[Arvind Rajaraman @arvindr02]

@_shreya_s @GreylockVC @rocksetcloud @AbnormalSec @bedrocksec Huge congrats!!

[Arvind Rajaraman @arvindr02]

@SuvanshSanjeev congrats!

[Arvind Rajaraman @arvindr02]

which bevi intern cooked

[Arvind Rajaraman @arvindr02]

@amydeng_ congrats!!

[Amit Kumar @amitku]

A simple idea to build the @UCBerkeley startup alumni network has grown beyond my wildest dreams into #AccelScholars, a tight-knit community of the most ambitious, talented, kind-hearted people, whose individual stories we’ve been fortunate to support for the past eight years

[Parth Asawa @pgasawa]

Instruct-tuned models are getting better at following instructions and ‘reasoning’ every day, but they’re shockingly poor at generating diverse responses. Diversity is crucial to many tasks like synthetic data generation. We tackle this with a new approach, BARE 🐻! (1/n)