About Me

I am Mingxuan Li, currently pursuing my PhD at the University of Pittsburgh, where my research is deeply rooted in applying AI to scientific discovery, particularly focusing on AI’s transformative potential in digital advancements. My work primarily explores the intersection between AI and optics design, with significant emphasis on computational electromagnetics and Bayesian Optimization.

As the project leader at the MDS-Rely center, I oversee an innovative project titled “Machine Learning for EMI Shielding Design,” aiming to leverage machine learning techniques to enhance electromagnetic interference shielding. Additionally, I contribute as a peer reviewer for IEEE journals.

My academic journey began with a Bachelor’s degree in Chemistry, during which I gained valuable experience designing and synthesizing small molecules for pharmaceutical applications. This experience has fostered a continued interest in cheminformatics, bridging my past and present scientific explorations.

Recently, I have ventured into developing open-source software, spearheading the JAX-based efficient transfer-matrix method framework, JaxLayerLumos. This project stands out by delivering performance that significantly surpasses conventional commercial solvers, offering a solution that is both more accessible and efficient. This achievement underscores the project’s innovative approach and its contribution to accelerating scientific computation.

Additionally, my recent internship as a Machine Learning Engineer Intern at Schrödinger has enriched my skills in cloud-native MLOps, enhancing our AutoML library’s training and inference speeds and contributing to the development of profiling and benchmarking tools.

In summary, my academic and research activities are guided by a commitment to excellence in AI research, with a specific focus on its application to optics design and computational sciences. I am dedicated to contributing to the advancement of this field, seeking out new challenges and opportunities for innovation.

News

November 6, 2024
I’m thrilled to share that our paper, “Discovering Multi-Layer Films for Electromagnetic Interference Shielding and Passive Cooling with Multi-Objective Active Learning,” has been accepted by the 38th Conference on Neural Information Processing Systems (NeurIPS 2024) AI4Mat workshop. I look forward to presenting our work at the conference in Vancouver and engaging with the research community.

October 20, 2024
The JaxLayerLumos community continues to expand! We have launched two new packages: JaxColor, which facilitates structural color simulations, and JaxLayerRF, designed for analyzing layer properties in the RF region. These additions empower users to explore broader spectral ranges and more complex optical behaviors.

September 1, 2024
My summer internship as a Machine Learning Engineer Intern at Schrödinger has just wrapped up, filled with immense learning and contributions, especially in enhancing our AutoML library’s training and inference speeds. Immense thanks to the entire ML team for their invaluable mentorship. Excited to leverage this enriched cloud-native MLOps experience in future machine learning ventures!

June 15, 2024
We are excited to announce the first major milestone for JaxLayerLumos, our JAX-based framework for optical simulations, which offers a lightweight, flexible, and fast alternative to traditional software, enabling complex simulations with features like gradient calculations and angled incidence support, now available for the optics and photonics community.

March 15, 2024
Exciting times ahead as the LayerLumos project celebrates hitting its first major milestone, earning rave reviews from the open-source community. Here’s to a year brimming with innovation and fruitful collaboration!

February 9, 2024
Thrilled to share that I’ve accepted an offer to embark on a new adventure as a Machine Learning Engineer intern at Schrödinger. Can’t wait to dive into this journey and explore what’s in store!

December 13, 2023
Proud moment for our team as we showcased our work on Datasets and Benchmarks for Nanophotonic Structure and Parametric Design Simulations at NeurIPS 2023. A huge shoutout to all our collaborators for making this possible!