Hamed Mahdavi

Pennsylvania State University, State College, PA

hmm5834 at psu.edu

I am a Ph.D. candidate in Computer Science and Engineering at Pennsylvania State University, advised by Prof. Vasant Honavar, graduating in November 2026. My research is on reasoning in large language models: I build agentic workflows and evaluation harnesses that find where models fail, generate targeted synthetic data to fix those failures, and feed it back through SFT, RL, and pretraining.

Most recently I was a Research Scientist Intern at a stealth AI startup in New York, training reward models for creative judgment. Before that I collaborated with Google's Gemini Data team on the Gemini IMO 2025 project, and my work on LLMs for mathematics is supported by an unrestricted research gift from Google.

I am on the job market for research scientist roles starting late 2026, with a focus on LLM post-training, reasoning, and data. For more information, you can check my CV, or reach me at hmm5834@psu.edu.

Work Experience

Research Scientist Intern

Stealth AI Startup - New York, NY
  • Developed agentic workflows and evaluation harnesses for open-ended creative content generation, with a focus on generating synthetic reasoning data for creativity.
  • Built pipelines that transform agentic workflows into training data, studying the interaction between mid-training and post-training for creative reasoning capabilities.
  • Trained reasoning and non-reasoning reward models for creative judgment and led data curation for reward modeling in the creative judgment domain.
Mar 2026 - Aug 2026

Collaborator, Gemini IMO 2025 Project

Google
  • Provided high-quality, verifiable data and evaluations to the Gemini Data team under the supervision of Dr. Vahab Mirrokni, resulting in a gold medal at the International Mathematical Olympiad (IMO).
Summer 2025

Research Assistant

Pennsylvania State University - State College, PA
  • Design agentic workflows for math problem solving and high-quality mathematical data generation; build and run evaluation harnesses for contest-level math (benchmarks, metrics, ablations, failure analysis).
Jan 2022 - Present

Data Scientist Intern

Arconic Technology Center - New Kensington, PA
  • Worked in the Fabrication Technology group under the supervision of Martin Marinack; tackled several data-driven optimization problems to improve Arconic's fabrication processes (welding, rolling, and casting).
Jun 2023 - Aug 2023

Research Intern

Max Planck Institute for Software Systems - Saarbrücken, Germany
  • Worked on fundamental aspects of reinforcement learning, focusing on multi-agent, adversarial, and human-centered reinforcement learning.
  • Studied how to design an AI agent that can robustly cooperate with agents of unknown types. The project was inspired by real-world applications in which an AI agent, e.g., a virtual assistant, must cooperate with new agent types/users after deployment.
Aug 2018 - Jan 2019

Software Engineering Intern

Faragostar - Tehran, Iran
  • Developed a compression algorithm that allowed Faragostar to retain its official documentation while minimizing storage requirements.
  • Configured infrastructure for Elasticsearch, enabling more efficient and complex searches in the company's app compared to the previous SQL-based search tool.
Jul 2017 - Sep 2017

Mathematical Olympiad Teacher

Allameh Helli High School - Tehran, Iran
  • Taught number theory to academically gifted students at Allameh Helli High School.
Oct 2012 - Apr 2013

Selected Publications

  • Splicing Hints into Reasoning Traces: Distillation Beyond What the Demonstrator Can Solve
    H. Mahdavi, P. Mahdavinia, M. Farhadi, V. Honavar
    Submitted to ICLR, 2027 · arXiv
  • RefGrader: Automated Grading of Mathematical Competition Proofs using Agentic Workflows
    H. Mahdavi, A. Hashemi, M. Daliri, P. Mohammadipour, A. Farhadi, S. Malek, Y. Yazdanifard, N. Mireshghallah, A. Khasahmadi, V. Honavar
    Math-AI Workshop at NeurIPS, 2025 · arXiv
  • CombiGraph-Vis: A Multimodal Olympiad Benchmark for Discrete Mathematical Reasoning
    H. Mahdavi, A. Hashemi, M. Daliri, P. Mohammadipour, A. Farhadi, S. Malek, Y. Yazdanifard, A. Khasahmadi, V. Honavar
    Math-AI Workshop at NeurIPS, 2025 · arXiv · Dataset
  • Harnessing Optimization Dynamics for Curvature-Informed Model Merging
    P. Mahdavinia, H. Mahdavi, N. Mireshghallah, M. Mahdavi
    Preprint, 2025 · arXiv
  • Brains vs. Bytes: Evaluating LLM Proficiency in Olympiad Mathematics
    H. Mahdavi, A. Hashemi, M. Daliri, P. Mohammadipour, A. Farhadi, S. Malek, Y. Yazdanifard, A. Khasahmadi, V. Honavar
    COLM Conference, 2025 · arXiv
  • Beyond Training Data: How Elemental Features Enhance ML-Based Formation Energy Predictions
    H. Mahdavi, V. Honavar, D. Morgan
    Digital Discovery, 2025 · Paper
  • Towards deployment of robust cooperative AI agents: An algorithmic framework for learning adaptive policies
    A. Ghosh, S. Tschiatschek, H. Mahdavi, A. Singla
    AAMAS Conference, 2020 · arXiv

Education

Pennsylvania State University

Ph.D. Candidate
Computer Science and Engineering
Advisor: Prof. Vasant Honavar · Thesis: Solving Contest-Level Math Problems Using LLMs
January 2022 - November 2026 (expected)

Sharif University of Technology

Master of Science
Computer Engineering - Artificial Intelligence
September 2017 - January 2021

Sharif University of Technology

Bachelor of Science
Computer Engineering
September 2012 - August 2017

Honors and Awards

$100k Unrestricted Research Gift from Google

Awarded to support research on enhancing large language models for mathematics through synthetic data generation and agentic workflows.
May 2025

Top National Rankings in M.Sc. Entrance Exam

Placed 5th (AI), 2nd (Networks and Security), and 4th (IT) nationwide.
June 2017

Ranked 217th in National University Entrance Exam

Top 0.1% among 260,000+ examinees nationwide.
June 2012

Silver Medal, 29th National Mathematical Olympiad

Selected as one of 40 participants from 100,000+ candidates.
September 2011

Interests

My interests include but not limited to:

  • LLMs for Math Problem Solving
  • AI Agents for Synthetic Data Generation
  • Reasoning in LLMs
  • Post-training, Reward Modeling, and LLM-as-a-Judge
  • Reinforcement Learning
  • Optimization in Machine Learning

In general, I'm interested in applied machine learning fields where it's essential to lay a solid theoretical foundation alongside practical application

Besides my interest in science, I also enjoy playing tennis, and badminton, which I've been doing since I was 9 years old. I like playing guitar too, both acoustic and electric, and I usually go for classic rock songs.