Hello, I'm Kiarash

I study how agents learn reusable structure, like skills, programs and models of the world.

I’m a PhD student in Computing Science at the University of Alberta, advised by Levi Lelis. I work in reinforcement learning, on problems where structure helps an agent generalize: discovering options for hierarchical RL, representing policies as programs, and using LLMs as world models.

Before my PhD I was a support researcher at Huawei in Edmonton for two years, working on image-quality assessment, ISP and compiler optimization. I did my MSc at UAlberta with Martin Müller, where I studied Monte Carlo Tree Search with imperfect models, and my BSc in Computer Engineering at Iran University of Science and Technology.

LLM world modelingReinforcement learningProgrammatic policy synthesisOption discovery & hierarchical RL

Download CV Get in touch

News

2026Didec (option discovery by decomposing neural policies) accepted at NeurIPS 2026.
2026Our study of OOD generalization in programmatic RL accepted at ICML 2026.
2026ScoreNet accepted at WACV 2026.
2024Started my PhD at the University of Alberta with Levi Lelis, and received the department’s Recruitment Award.
2024UA-MCTS, from my MSc thesis, published at AAAI 2024.