Mingwan Zhang, Ph.D. Email and Phone Number

• Investigated recommendation systems by various matrix factorization methods by adding or not adding bias/side information of customers; results have similar runtime and memory usages with increasing 3% in precision and decreasing 11% - 15% in RMSE and MAE• Predicted the customer behaviors for product sales on Instacart virtual grocery for stocking optimization using timeseries machine learning algorithms during COVID-19; compare with long-short term memory (LSTM) network, graph convolutional network (GCN) lowers the 29% of the MSE• Improved the accuracy for distinguishing enzyme structures from non-enzymes by 12%, analyzing over a thousand protein structures using deep learning than SVM; ~5 times saving method by deep learning because of less used features• Predicted magnetic interactions between a pair of atoms in molecules by machine learning in different models: Neural Network, Gradient Boost Tree, Random Forest; conclude tree-based ensemble models have better results than that of the neural network models by decreasing the ~50 % RMSLE, and Random Forest performs even better by furtherdecreasing RMSLE ~6%• Lowered at least 10-times cost than traditional experimental method for predicting transparent conductors by GaussianProcess Regression, which is a better machine learning algo than SVM by decreasing MSE to 67%

• Courses: Introduction to Computing for Engineers (Python, all level), General Physics II (engineering students) et al.• Delivered knowledges to thousands of different levels of learner effectively with overall 90%+ positive feedback• Introduced new system and guidance rules for courses and laboratories teaching to improve efficiency of instructorsand students

• Collaborated with multiple National Laboratories to drive research development from scientific scenario to achievement• Scaled up modeling system by ~8 times capability for lab synthesized semiconductor materials in various applications• Applied Python and MATLAB to statistical demonstration and visualization for millions of lab measured and testingsemiconductor sensors signal results and catalytic performances