Hamed Khosravi

I am a Machine Learning Ph.D. student at Georgia Tech, fortunate to be advised by Prof. Xiaoming Huo.

My research focuses on developing statistical machine learning, causal inference, and reinforcement learning methods for evaluating, understanding, and improving AI systems, with an emphasis on large language models and sequential decision-making systems. I also spent 2025 at Tesla building production ML pipelines for large-scale modeling and evaluation.

I am always glad to talk about research or possible collaborations — feel free to reach out.

Research Interests

• Statistical machine learning
• Reinforcement learning
• LLM evaluation and risk control
• Reinforcement learning with verifiable rewards (RLVR) for LLMs
• Human–AI teaming
• Statistical evaluation of AI systems
• Multi-armed and contextual bandits
• Non-stationary and online learning
• Sequential decision making
• Conformal prediction and uncertainty quantification
• Causal inference and experimental design

Latest Papers

A selection — all publications →

1. Conformal Selective Acting: Anytime-Valid Risk Control for RLVR-Trained LLMs Hamed Khosravi, Xiaoming Huo NeurIPS 2026 (main track) Paper Code
2. Which LLM for Which Work? Budgeted Model Allocation under Uncertain Evaluation Hamed Khosravi, Xiaoming Huo Can We Trust AI Evaluation? Workshop (TAE), NeurIPS 2026 Paper Code
3. Beyond “AI Helps Humans”: Decision-Targeted Evaluation Design for Human–Agent Teams in the Agentic Era Hamed Khosravi, Xiaoming Huo Human–AI Coevolution Workshop (HAIC), NeurIPS 2026 Paper Code
4. LLM-Observed Multi-Agent Maintenance with Amortized Reinforcement Learning Policies Hamed Khosravi, Xiaoming Huo SSRN preprint, 2026 Paper Code
5. Catching a Moving Subspace: Low-Rank Bandits Beyond Stationarity Hamed Khosravi, Xiaoming Huo arXiv preprint arXiv:2605.20269, 2026 Paper Code

Experience

• Jan – Aug 2025 Machine Learning Engineer Intern, Tesla — Cell Engineering, Analytics & Software
  • Developed a modular Python library for automated machine learning workflows, integrating data preprocessing, model selection, interpretability, and reporting. Built to scale within the analytical ecosystem for fully automated classification, regression, and time series modeling.
  • Designed and deployed ML pipelines on Kubeflow, using its Kubernetes-based infrastructure to run and manage containerized steps for scalable, reliable model training and evaluation.
  • Integrated conformal prediction and contrastive learning into analytics workflows to improve anomaly scoring reliability and decision support.
  • Applied machine learning to real-world manufacturing data, collaborating with software and quality teams to balance model performance, interpretability, and practical constraints.

Education

• 2025 – present Ph.D. in Machine Learning, Georgia Institute of Technology GPA 4.0/4.0 · Advisor: Prof. Xiaoming Huo · H. Milton Stewart School of Industrial & Systems Engineering
• 2025 M.S. in Industrial Engineering (Data Science), West Virginia University GPA 4.0/4.0 · Advisor: Dr. Imtiaz Ahmed · Co-advisor: Dr. Srinjoy Das · Department of Industrial and Management Systems Engineering
• 2022 B.S. in Industrial Engineering, Iran University of Science and Technology, Tehran, Iran