Edgar Hildebrandt Rojo | Applied ML & Algorithmic Systems

Applied ML & Decision Systems

I build machine learning systems that evaluate and optimize decisions under uncertainty.

Right now, most of my work sits at the intersection of algorithmic optimization, distributed simulation, and LLM evaluation. I don’t just focus on training models; I spend my time engineering the systems around the model: the environments, objective functions, and evaluation frameworks that dictate whether an algorithm actually works in the real world.

At Johns Hopkins Applied Physics Laboratory, I build large-scale simulation and ML pipelines that drive operational decisions for the U.S. Navy and Coast Guard.

selfie upscaled.png

My core focus includes:

Distributed Simulation: Architecting Monte Carlo pipelines scaling to 1,000,000+ episodes on HPC clusters to evaluate policy trade-offs under stochastic conditions.
Algorithmic Optimization & RL: Building closed-loop architectures and training optimization agents (PPO/SAC) to dynamically adapt policies for constrained resource allocation and autonomous sensing.
LLM Evaluation: Developing domain-specific evaluation frameworks and schema compliance protocols to rigorously measure hallucination and logical robustness.

An Interdisciplinary Approach

I am fundamentally an interdisciplinary problem solver. My academic background, a triple major in EECS, Applied Math, and Economics from UC Berkeley (Go Bears! 🐻), followed by an M.S. in AI from Johns Hopkins, gives me a unique lens on machine learning.

I don’t just focus on pure theory, and I don’t just write code for deployment. I use my foundation in econometrics and applied mathematics to bridge the two. This allows me to define measurable objectives, model complex trade-offs, and ensure the systems I build actually solve the underlying structural problems.

While I currently apply this toolkit to defense and autonomy programs, the core mathematics of my work are entirely domain-agnostic. Whether I am modeling physical logistics or designing data-driven product architectures, my goal is the same: integrating rigorous, multi-disciplinary math with software engineering to build high-agency systems that drive real-world impact.

Built with Nooxy