Rati Devidze

I am a Reinforcement Learning & Machine Learning Engineer at minds.ai focused on applying Reinforcement Learning (RL) and Preference Learning to semiconductor fabrication (FAB) optimization.

I completed my PhD in Machine Teaching group of Max Planck Institute for Software Systems: MPI-SWS, in Saarbrücken, Germany. I was also a member of the Saarbrücken Graduate School of Computer Science. My doctoral research was supervised by Prof. Adish Singla.

My work centers on developing principled learning frameworks for complex, stochastic, and highly constrained environments. In particular, I study reward design aiming to construct informative and interpretable signals that guide agent behavior toward optimal decisions. These approaches enable efficient learning in settings where explicit reward specification is challenging, while aligning agent behavior with domain-specific objectives such as throughput maximization, cycle time reduction, and resource efficiency.

More broadly, my research interests include Reinforcement Learning and its subfields—such as reward design, inverse reinforcement learning, and meta-reinforcement learning as well as preference learning for sequential decision-making.

For more information about me, please find my [C.V.].