An AI researcher, who was hired by Meta’s Superintelligence Lab, has called it quits months after the Mark Zuckerberg-owned company hired him at a million-dollar salary. Rishabh Agarwal announced that he is on notice period on social media, adding that he “really enjoyed” working with the team. 

‘It was a tough decision’

“This is my last week at AI at Meta,” Agarwal announced on X (formerly Twitter), before adding that it was a “tough decision” to make. “It was a tough decision not to continue with the new Superintelligence TBD lab, especially given the talent and compute density.”

He added that he is planning to take a “different kind of risk” after spending 7.5 years at Google Brain, DeepMind, and Meta. 

ALSO READ

Although Agarwal said that the pitch from Zuckerberg and Alexandr Wang, who is Chief AI Officer at Meta, was great, he took their “compelling” words to heart. 

‘Biggest risk you can take is…’

“But I ultimately choose to follow Mark’s own advice: ‘In a world that’s changing so fast, the biggest risk you can take is not taking any risk’,” he said.

In the next few lines, he highlighted how he was able to contribute during his short stint at the company. “In my short time at Meta, we did push the frontier on post-training for ‘thinking’ models. Specifically, pushing an 8B dense model to near Deepseek-R1 performance with RL scaling, using synthetic data mid-training to warm-start RL, and developing better on-policy distillation methods.”  

Who is Rishabh Agarwal?

Rishabh Agarwal is an AI researcher and an alumnus of IIT Bombay, where he studied Computer Science and Engineering. His undergraduate research focused on “approximating the evaluation function in Scrabble”.

He went on to pursue his PhD in Artificial Intelligence at Mila – Quebec Artificial Intelligence Institute in Canada. 

Agarwal started his career as an intern with Saavn’s Search and Algorithms division, and then interned at Tower Research Capital’s Algorithm trading division. He was also a research intern at Waymo. 

ALSO READ

In June 2018, he joined Google Brain as a Senior Research Scientist and worked there for five years, advancing deep reinforcement learning (Deep RL).

“Won the NeurIPS 2021 best paper award for better evaluation in RL. Popularised offline deep RL (An optimistic perspective on offline RL, RL Unplugged). Achieved human-level performance on Atari with Human-level efficiency,” were some of his contributions he listed on LinkedIn. 

Following this, he joined Google’s DeepMind and spent two years working on large language models (LLMs) using reinforcement learning (RL), self-improvement, and distillation. 

In April this year, he joined Meta Superintelligence Labs as a Research Scientist. 

Alongside his research roles, Agarwal has also been involved in academia, teaching at McGill University as an Adjunct Professor.