Shayan Talaei
Ph.D. Candidate, Stanford University
LLM Researcher, Ricursive Intelligence
Advisors
Amin Saberi, Azalia Mirhoseini
Current Mentees
Alexander Huang, Kaitlyn Wang, Gino Chiaranaipanich, Amir Zeinali
Past Mentees
Emil Biju (Microsoft), Zhemin Huang (Microsoft AI), Meijin Li (Meta), Kanu Grover (Databricks), Ajay Anubolu (Together AI), Abhinav Chinta (Stanford), Agam Bhatia (Stanford)
I am a Ph.D. student at Stanford University, advised by Azalia Mirhoseini (Scaling Intelligence Lab) and Amin Saberi (Language Data and Reasoning Lab). I am also an LLM researcher at Ricursive Intelligence.
My research centers on how language models learn and reason. Today's models are data-hungry and unable to keep learning after deployment; I work toward closing this gap along three threads: test-time scaling and multi-agent systems that extend what models can achieve at inference; efficient, long-horizon reasoning that enables models to plan and parallelize their thinking; and post-training methods through which models internalize new knowledge via self-distillation. If you see an overlap of interests, feel free to send me an email!
Prior to Stanford, I obtained my B.S. in Computer Engineering from Sharif University of Technology, and had the opportunity to work on distributed optimization supervised by Dan Alistarh at ISTA, and on learning theory supervised by Emmanuel Abbe at EPFL.
Outside of research, I enjoy reading books, playing table tennis, and sometimes solving and designing Mathematical Olympiad problems.

Selected Publications

* denotes equal contribution

Awards