arun rao

linkedin twitter github

i help the NVIDIA Nemotron team build open-source large language models (LLMs) that any person or organization in the world can use and modify for their needs. i’m also affiliated faculty at UCLA, teaching a class on advanced product design, and am a board member of Positive AI Labs, helping figure out how to align AI models and systems for human flourishing.

previously at Meta, i played a lot with foundation agents, trying to perfect them to launch the little fellas at billion scale. i built out the evaluations org for Meta’s Llama autoregressive LLMs in Meta’s Central GenAI Group (a sister team to FAIR, now called Meta Superintelligence Labs, aka MSL). and i co-led a large team of product managers and research engineers working on ML recommenders and ranking models at Meta’s Monetization Ranking AI group that served ~4 billion people.

i spend my time thinking about real-world uses of machine learning, obscure math, internet protocols, APIs, incentive mechanisms, contemporary art, and design.

i write the Substack letter Hash Collision, where i report on developments in AI/ML, science, tech, culture, economics, and ethics. i co-wrote “A History of Silicon Valley, 2nd Ed.” i also co-host a podcast where i interview high variance, creative people, called the Variance Podcast, and support the Accelerated Math and Physics (AMP) Program for talented children in the Bay Area.

in the past, i led an applications and platform machine learning team at Amazon Music in San Francisco. prior to that, i was a co-founder of Starbutter AI (Skydeck F17) and a quant debt and derivatives trader at PIMCO (working under Bill Gross, Mohamed El-Erian, and Dan Ivascyn). i’m still a licensed attorney in the Wyoming State Bar (the Equality state, where women first got the right to vote, and LLCs and DAOs to exist) and was formerly a CFA charterholder.

i’m a grad of UCLA, Penn/Wharton, the University of Arizona College of Law, and the Indiana Academy.

i like books, running, yoga, math, grammar and style, dead ancient languages, modern computer languages, black hole physics, bears, cacti, and pad see-ew.