My name is Natalia A Rodriguez Figueroa and I am a born and raised Puerto Rican from Mayagüez, Puerto Rico. My academic background includes a Bachelor’s (BS) Degree in Industrial Engineering, with a minor in Project Management, from the University of Puerto Rico at Mayagüez. Most recently, I earned a Master’s (MS) Degree in Operations Research from University of California, Berkeley.
You can contact me through my current academic email: natalia_rodriguezuc@berkeley.edu
You can also find me on LinkedIn here:
I am interested in machine learning, data science, and their growing range of real-world applications, particularly in fields like healthcare, engineering systems, and business operations. Even so, what excites me most is going beyond implementation: using optimization, mathematical modeling, and statistical learning to improve how intelligent systems are built and understood.
While my foundation is in operations research and machine learning, I am increasingly focused on how optimization helps us design smarter, more interpretable, and more efficient AI systems. I’m especially drawn to the theoretical contributions of machine learning, and how rigorous modeling can support advances in algorithmic performance and real-world impact. I'm currently learning more about AI structures through advances on my past and current project applications.
I currently have two main projects that I worked on during my undergraduate and graduate journey, which I have continued to build on since then. At UPR Mayagüez, my work centered on applying machine learning frameworks to improve survival prediction in pancreatic cancer, using feature selection techniques to enhance model performance. Later at UC Berkeley, I developed optimization methods to improve training efficiency in contrastive learning tasks, focusing on proximal and Newton-based approaches for AUC maximization.
I recently completed a Computing Graduate Internship at Lawrence Livermore National Laboratory (LLNL) through the Computing Scholar Program. My project focused on developing computationally efficient algorithms for global optimization of acquisition functions in Bayesian Optimization, particularly under high-dimensional, high-fidelity data settings.
Building on this experience, I am continuing to support LLNL’s HiOp open-source repository, contributing research efforts on acquisition optimization problems using branch-and-bound methods. Our goal is to advance scalable and reliable strategies for global optimization of black-box functions, with applications in Bayesian Optimization and beyond.
|
|
|