Researcher at SLATE

Sam Urmian

My research develops algorithmic and machine learning foundations for trustworthy AI and data science. I study adaptive AI systems that learn from data, feedback, and human interaction, with a focus on the complexity and reliability of learning under real-world constraints.

A central theme in my work is feedback complexity: understanding what resources—data, computation, communication, verification, and interaction—are needed when a model’s outputs influence the future data it observes. This perspective connects problems in synthetic data generation, personalized recommendation, federated learning, learning-guided reasoning, automated theorem proving, and ML-assisted combinatorial optimization.

Across these projects, my goal is to design algorithms that are reliable, interpretable, privacy-preserving, personalized, and computationally efficient. On the application side, I apply these ideas to educational technology and learning analytics, especially in settings where AI systems must adapt to learners while maintaining trust, privacy, and robustness.

Currently, I am a researcher at the Centre for the Science of Learning & Technology (SLATE), working with Mohammad Khalil and Barbara Wasson. I will defend my PhD in August 2026 under the supervision of Mateus de Oliveira Oliveira on dynamic programming algorithms and automated theorem provers, at the Department of Informatics, University of Bergen. Before that, I received my master's degree from Sharif University of Technology.