Giving up a million-dollar pay package and a position in the prestigious Superintelligence team at Meta may sound ridiculous. However, Rishabh Agarwal, a computer science graduate from IIT Mumbai who also obtained a doctorate in computer science from Mila-Quebec AI Institute, did just that.
Agarwal put in his papers in favour of exploring other challenges after being part of Meta AI’s Superintelligence team for barely five months. Why? Well, after much consideration, he thought it best to follow Mark Zuckerberg’s advice that the biggest risk in today’s rapidly-evolving world is not to take any risk at all! However, Agarwal is yet to reveal what “different kind of risk” he wishes to pursue.
As of now, his decision to let go a seven-figure package in a bid to seek other challenges has created ripples on social media. He has left netizens wondering why anyone would quit a place in the Superintelligence team put together by Mark Zuckerberg, CEO, Meta and Alexander Wang, chief AI officer, Meta. The team comprises the best talent drawn by the company with lucrative packages, some of whom have been poached from competition.
During his short stint at the organisation, Rishabh Agarwal managed to use RL scaling to push an 8-bit dense model to match the performance of Deepseek R1. He also he used synthetic data in mid-training to warm-start RL and assisted in the development of improved on-policy distillation methods.
Agarwal has interned with names such as Waymo, Saavn, and Tower Research Capital. He has also served Google Brain in a full-time role as a senior research scientist. He then went on to take up the role of staff research scientist at Google DeepMind before joining Meta Superintelligence Labs as a research scientist.
It is pertinent to mention here that Agarwal is not the only one to take such a decision. About three researchers have reportedly quit the team in recent times. What’s up with this ambitious team? Speculations are on.
