Your Photo

Dr. Luka Grbcic

Computational Researcher | Machine Learning and Optimization in Science and Engineering

Email GitHub LinkedIn

Bio

Welcome to my personal page! I am a postdoctoral researcher at the Lawrence Berkeley National Laboratory in the Applied Math and Computational Research Division. I am a part of the Applied Computing for Scientific Discovery group, where my research is focused on developing computational tools and methods that accelerate scientific discovery. More specifically, my research work encompasses topics such as hybrid machine learning and optimization methods, batch active learning for autonomous experimentation, tandem neural networks, and surrogate black-box optimization. Currently also interested in developing LLM agents and evolutionary inspired frameworks for algorithm improvement. All of my research motivation and ideas are driven by real-world problems in science and engineering. Based in the SF Bay Area.

Research Statement

My research vision is dedicated to accelerating scientific discovery through the innovative integration of advanced computational methods, particularly at the intersection of machine learning and optimization. My work is driven by the imperative to solve complex, real-world problems, with a focus on developing robust and efficient tools for inverse design and autonomous experimentation.

A core theme of my research involves the development and application of hybrid machine learning and optimization methods. This includes pioneering techniques such as greedy surrogate-based optimization and multi-fidelity ensemble frameworks to efficiently navigate vast parameter spaces for complex systems, as demonstrated in my work on photonic surfaces. I am particularly interested in batch active learning for autonomous experimentation, enabling intelligent decision-making and optimal resource utilization in scientific workflows. I am also actively developing LLM agent frameworks, inspired by evolutionary algorithms, tailored for improving algorithms for active learning and optimization.

Furthermore, my expertise extends to the design and implementation of tandem neural networks for inverse design problems, providing powerful solutions for mapping desired outcomes back to their underlying causes. I also focus on surrogate black-box optimization, creating efficient approximations for computationally expensive simulations to enable faster and more effective design cycles.

My ultimate goal is to bridge the gap between theoretical advancements in computational science and their practical application, creating impactful solutions that push the boundaries of scientific understanding and technological innovation. I am always seeking interdisciplinary collaborations to apply these methodologies to new and challenging domains.

Experience

Postdoctoral Researcher

Lawrence Berkeley National Laboratory, Berkeley, California, USA

January 2023 – Present

  • Conducted research on machine learning methods for engineering and science.
  • Developed and implemented tailored solutions for autonomous experimentation.
  • Collaborated with a variety of domain scientists on utilizing AI to accelerate experimentation.
  • Developed software for solving inverse design problems with Tandem Neural Networks.

Doctoral Researcher / Research and Teaching Assistant

University of Rijeka, Rijeka, Croatia, EU

October 2016 – September 2021

  • Ph.D in Fundamental Technical Sciences - Computational Mechanics.
  • Completed Ph.D. thesis titled "Numerical modeling of fluid mixing in pipe networks with machine learning applications".
  • Published 23 papers in peer-reviewed journals.

Latest Publications

All publications available at: Google Scholar Profile

Conference Talks

Recorded Talks