Xiangyu Yin

Xiangyu Yin

I am a Postdoctoral Appointee at the Argonne National Laboratory. My research interests include AI4science, Physics4ML, and autonomous scientific discovery. My overall research themes include integrating physics with data and algorithms, improving scientific discovery with new methodologies, as well as bridging gaps between different scientific domains. My typical work involves communicating with scientists and engineers, defining and analyzing various scientific and engineering problems, proposing, and implementing solutions by researching the latest technologies in optimization, statistics, data science, machine learning, and software development, and finally reporting and publishing results as well as following up with stakeholders and executors. Before joining Argonne, I obtained my Ph.D. in Chemical Engineering from Carnegie Mellon University, where I carried out research in computational materials discovery.

Experience

 
 
 
 
 
Postdoctoral Researcher
Argonne National Laboratory
Oct 2023 – Present Lemont IL
  • Working across various projects under the common theme of automatic parameter tuning for computational imaging technologies
 
 
 
 
 
Research Intern
Toyota Research Institute
May 2022 – Aug 2022 Los Altos CA
  • Proposed a reinforcement learning (RL) framework for designing dopants and demonstrated its efficiency and scalability
  • Developed a package for RL-based dopant design, including customized environment, agent, and trainer components
  • Conducted experimental data analysis to develop predictive models, performed explainability studies, and optimized properties
 
 
 
 
 
Graduate Researcher
Carnegie Mellon University
Sep 2018 – Sep 2023 Pittsburgh PA
  • Coursework: Optimization, Statistics, Machine Learning, MLOps, Deep Learning
  • Teaching Assistant: Chemical Product Design, Introduction to Chemical Engineering
  • Award: Neil and Jo Bushnll Fellowship in Engineering

Research

Process-Materials Integrated System Modeling & Co-optimization
Advanced integrated energy systems consist of components across a wide range of scales. We aim to develop methodologies that bridge the gap between different scales, starting from simultaneous optimization of process and materials.
Last updated on May 8, 2023
Process-Materials Integrated System Modeling & Co-optimization
Advanced Optimization Strategies for Accelerated Materials Discovery
Optimization is at the core of accelerated materials discovery. We aim to develop advanced optimization strategies that are accurate, efficient, scalable, interpretable and generalizable for computational materials design problems.
Last updated on Jun 28, 2023
Advanced Optimization Strategies for Accelerated Materials Discovery

Projects

Deep Reinforcement Learning for Dopants Design
Movie Recommendation Service System
ML-assisted Vehicle Routing Under Uncertainty
PyTorch-like Deep Learning Framework & NeuralODE
Speech Recognition System & Single Image Super-Resolution
Million Song Dataset Analytics
Lot-streaming Flowshop Scheduling Optimization
Double Traveling Salesman Problem Optimization

Publications

Xiangyu Yin, Lorenz Biegler, Chrysanthos E. Gounaris (2023). Inverse Design of Metal-Organic Frameworks for Adsorption Processes: Learning Numerically Optimizable Surrogate Models of Isotherm Parameters.. In preparation.

Xiangyu Yin, Chrysanthos E. Gounaris (2022). Computational Discovery of Metal-organic Frameworks for Sustainable Energy Systems: Open Challenges. In Comput. Chem. Eng..

PDF Cite DOI

Christopher L. Hanselman, Xiangyu Yin, David C. Miller, Chrysanthos E. Gounaris (2022). Matopt: A python package for nanomaterials design using discrete optimization. In J. Chem. Inf. Model..

PDF Cite Code DOI

Xiangyu Yin, Chrysanthos E. Gounaris (2021). Search methods for inorganic materials crystal structure prediction. In Curr. Opin. Chem. Eng..

PDF Cite DOI

Xiangyu Yin, Natalie M. Isenberg, Christopher L. Hanselman, James R. Dean, Giannis Mpourmpakis, Chrysanthos E. Gounaris (2021). Designing stable bimetallic nanoclusters via an iterative two-step optimization approach. In Mol. Syst. Des. Eng..

PDF Cite DOI

Skills

Programming: Python, C++, SQL, MATLAB, Shell

Libs/Frameworks: Numpy, SciPy, SciKit-learn, PySpark, Ray, CUDA, PyTorch, TensorFlow, JAX, Gymnasium, Hugging Face, LangChain, AutoGluon, Docker, Pyomo, Gurobi, IPOPT, BARON, NOMAD, RASPA

Platforms: Github, AWS, Azure, GCP(Google Data Analytics), OpenAI

Talks

X. Yin, R. Lei, W. Ye, and J. Montoya. Deep Reinforcement Learning for Dopants Design in Crystalline Materials. MRS Fall Meeting, 2023

X. Yin, L. T. Biegler, and C. E. Gounaris. Incorporating Materials Surrogate Models into Process Models for Adsorption-Based Gas Separations. AIChE Annual Meeting, 2022

X. Yin, L. T. Biegler, and C. E. Gounaris. Enabling Process-materials Co-optimization via Surrogate Modeling for Adsorption-Based Gas Separations. INFORMS Annual Meeting, 2022

X. Yin, L. T. Biegler, and C. E. Gounaris. Towards process-materials co-optimization: Automatic generation of optimizable structure-function relationships. AIChE Annual Meeting, 2021

X. Yin, C. L. Hanselman, and C. E. Gounaris. Designing stable semiconductor nanowires via a generic materials optimization toolkit. AIChE Annual Meeting, 2020

X. Yin, C. L. Hanselman, D. C. Miller, and C. E. Gounaris. Nanomaterials design using Pyomo. INFORMS Annual Meeting, 2020

X. Yin, C. L. Hanselman, D. C. Miller, and C. E. Gounaris. Matopt: A mathematical optimization-based nanomaterials design toolkit. INFORMS Annual Meeting, 2020

X. Yin, N. M. Isenberg, M. G. Taylor, G. Mpourmpakis, and C. E. Gounaris. Identification of stable bimetallic nanoclusters via a mathematical optimization framework. AIChE Annual Meeting, 2019

Contact