Keiran Paster

I am a fifth-year PhD candidate (on leave) in the Department of Computer Science at the University of Toronto and an early employee at xAI.

• I am fortunate to be advised by Jimmy Ba and Sheila McIlraith.
• I am also affiliated with the Vector Institute.
• In Fall 2023, I was a student researcher at Google working on data efficiency for LLMs with the Gemini and Blueshift teams.

Highlights from my time at xAI include:

• Grok 2 and Grok 2 mini, where I did the web data filtering, ablation, and data mixing efforts in the data team. I also contributed to the post-training of the “sus-column-r” model, xAI’s first model on Chatbot Arena.
• Grok 3 and Grok 3 mini, where I did the web data filtering, ablation, and data mixing. Additionally, I was an initial member of the reasoning team and a core contributor to the reasoning models, where I did the training and data recipes.
• Grok 4, where I led the RL data and ablation efforts as we scaled up our RL training by over 10x.
• Grok 4 Fast, Grok 4.1, Grok 4.1 Fast, and Grok 4.20, where I led the Grok Next team, overseeing the reasoning and post-training teams. Some highlights during this time are our advancements in post-training, search, mathematical proofs, forecasting, and multi-agent.