Amirhossein Afsharrad

PhD Candidate, Stanford University

Email: afsharrad@stanford.edu
Google Scholar
LinkedIn

*Note: You might see my name as either Amir or Amirhossein in different places. Both work! If you're confident in your pronunciation of the full version, go for it - otherwise, Amir is the easy button.

I am a PhD candidate in Electrical Engineering at Stanford University, advised by Professor Sanjay Lall.

My current research primarily focuses on post-training of large language models, LLM alignment with human preferences, and evaluation and reward modeling of LLMs.

I am also interested in more theoretical problems in optimization and reinforcement learning, including problems at the intersection of optimization and control such as applications of transformer models in control theory and autonomous systems, as well as bandit problems with safety constraints and multi-agent settings. These areas were a primary focus earlier in my PhD, and I continue to work on projects in these directions from time to time.

Prior to Stanford, I received my B.Sc. in Electrical Engineering and Computer Science from Sharif University of Technology, where I was advised by Prof. Mohammadali Maddah-Ali.

News

January 2026: Our paper “Beyond Binary Preferences: A Principled Framework for Reward Modeling with Ordinal Feedback” was accepted to ICLR 2026 [paper]
January 2026: Our paper “Multi-Agent Stage-wise Conservative Linear Bandits” was accepted to ACC 2026 [paper]
December 2025: Presented “LORE: Lagrangian-Optimized Robust Embeddings for Visual Encoders” at NeurIPS 2024 in San Diego [paper]