I build machine learning systems that evaluate and optimize decisions under uncertainty.
My work spans reinforcement learning, distributed Monte Carlo experimentation, and LLM evaluation, with a focus not just on training models, but on designing the environments, objectives, and evaluation frameworks that determine whether those models produce reliable outcomes in complex systems.
How do we design AI systems that make robust decisions when objectives conflict and uncertainty dominates?
At Johns Hopkins Applied Physics Laboratory, I design and deploy large-scale simulation and machine learning infrastructure supporting operational decision-making for the U.S. Navy and Coast Guard.
My core focus includes:
Distributed Experimentation & Simulation:
Architecting Monte Carlo pipelines scaling to 1,000,000+ episodes on HPC clusters to evaluate policy trade-offs under stochastic conditions.
Reinforcement Learning Systems:
Building closed-loop RL environments integrating real-time system state for constrained resource allocation and autonomous decision logic.
LLM Evaluation & Robustness:
Developing domain-specific evaluation frameworks and schema compliance protocols to measure hallucination and logical robustness.
Rather than focusing solely on model performance, I specialize in the systems around the model: objective functions, constraints, reward design, and evaluation protocols that determine whether an AI system survives contact with reality.
My approach is grounded in a triple major in EECS, Applied Mathematics, and Economics from University of California, Berkeley, followed by an M.S. in Artificial Intelligence from Johns Hopkins University. This blend of systems engineering, statistical rigor, and economic reasoning informs how I build practical, high-agency AI systems.
I’m interested in joining high-autonomy teams building AI systems that reason, plan, and optimize within complex physical and economic environments.
The same tools used to evaluate autonomous operational policies — large-scale experimentation, optimization under constraints, and robustness testing — generalize directly to economic and marketplace environments.
My academic foundation informs how I think about incentives, trade-offs, and measurable impact. Whether in physical or economic systems, the goal remains the same: design decision architectures that perform reliably under uncertainty.
© 2026 Edgar Hildebrandt Rojo
Algorithmically served by Nooxy