I am a PhD candidate at Washington State University, advised by Prof. Jana Doppa.
I study how AI systems can make good decisions when trial and error is expensive or risky. My research spans safe reinforcement learning, where agents must maximize performance while keeping costs and risks within limits, and offline model-based optimization, where the goal is to discover better designs using only data from past experiments. This work is motivated by high-stakes applications such as materials discovery, healthcare, and smart grids.
Agents that maximize reward while satisfying safety constraints, whether they learn from logged data (offline) or through interaction (online).
Discovering high-performing designs from a fixed dataset of past experiments, when every new evaluation is costly.
* equal contribution
I am always happy to discuss research and potential collaborations. You can reach me at yassine.chemingui [at] wsu.edu.