Xiaoting Zhong

Data Scientist @ Lam Research

About

Ph.D. researcher with 7+ years of experience applying Machine Learning and Data Analysis in cross-disciplinary studies. I analyze data of complicated nature and diverse formats to make accurate predictions. Technical skills and tools: Python, Pytorch, mpi4py, Ray Tune, SQL Some of my experiences include: 1) Predicted material mechanical strength from images with 12 % Mean Absolute Percentage Error (MAPE) using Deep Convolutional Neural Networks. 2) Achieved material instance segmentation with 86.64 % mean Average Precision (mAP) using only 20 training images and wrote a Web-based App to allow easy analysis of the model predictions. 3) Reviewed 42 eXplainable Artificial Intelligence (XAI) techniques for materials science applications. 4) Built an image-to-image translation surrogate model for 3D printing simulations with a 0.8 FID score. I'm passionate about applying Machine Learning and Data Analysis to generate perceptive and actionable insights from complicated data. Data tells its story, and I make the story understandable.

Country

United States

City

Livermore

Industry

Semiconductors

Skill

Deep Learning, OpenCV, PyTorch, mpi4py, High Performance Computing (HPC), Ray Tune, Regular Expressions, DASH, Plotly, Python (Programming Language), Keras, NumPy, Pandas, Scikit-Learn, SciPy, Matplotlib, Machine Learning, Data Analysis, Scanning Electron Microscopy, Metallurgy

Experience

Data Scientist

Lam Research

2022-9 - Present · 4 yrs 1 mo

Fremont, California, United States

Postdoctoral Research Fellow

Lawrence Livermore National Laboratory

2019-10 - 2023-1 · 3 yrs 4 mos

Livermore, California, United States

My research experience includes: ∙ Accelerate Additive Manufacturing process simulations by building Random Forest and conditional Generative Adversarial Network surrogates to perform image-to-image translations and achieved a 0.8 FID score. ∙ Review 42 eXplainable Artificial Intelligence (XAI) techniques with 15 recent materials science research examples. ∙ Segment and detect complicated microstructural features in TEM images using less than 20 training images and achieved 86.64% mean Average Precision (mAP). ∙ Predict the ultimate compressive strength of high-energy materials from (50,000+) experimental SEM images using Wide Res Net and Random Forest with VGG16 transfer learning features. Achieved 13 % Mean Absolute Percentage Error (MAPE). ∙ Study machine learning model robustness to real-world instrument-shift-induced data quality variations and evaluated data standardization methods to improve ML model robustness. ∙ Design Active Learning experiment planners for polymer electrolyte Additive Manufacturing process.

Graduate Research Assistant

Carnegie Mellon University

2014-8 - 2019-8 · 5 yrs 1 mo

Greater Pittsburgh Area

My research experience includes: • Reconstruct and clean nine different 3D microstructures from 500+ experimental orientation maps (total size 10+ GB). • Optimize the computation of grain boundary energy as an eigenvalue problem and accelerated the computation speed by over 100 times. • Designed an algorithm to compute the five-dimensional grain boundary curvature distribution within 3D microstructures. • Identified the bi-model distribution of grain boundaries and found the energy-curvature correlation of general boundaries and singular boundaries and opposite signs (- 0.38 and 0.57). • Measured the mobility of 20+ grain boundaries by conducting high-throughput Focused Ion Beam (FIB) experiments. • Track and analyze the evolution of 2000+ experimentally measured 3D Grain Faces.