Ryan Mowlai

I am a PhD candidate in Systems and Industrial Engineering at the University of Arizona, expected to graduate in May 2027. My dissertation develops reinforcement learning and agentic AI methods for autonomous operation of ultrafiltration water treatment systems, work funded by the U.S. Army Corps of Engineers. The models run on real plant data and real hardware, which shapes how I think about evaluation: a policy is only interesting if it holds up on the machine, not just on a benchmark.

My path into this started with prediction. At Sapienza I worked on transformer-based human pose forecasting for human robot collaboration. The PhD moved me from predicting what a system will do to deciding what it should do next.

Alongside the PhD I have worked at IBM on large-scale propensity modeling and LLM applications, at Lightsense Technology on multimodal deep learning and adversarial evaluation, and at DIDO Srl in Milan on PPO-based building energy optimization deployed on AWS. I have published at IEEE SIEDS, with additional work in progress.