Min Lin

Staff Machine Learning and Simulation Engineer

Western Digital

Jul 2023 - Current (3 years 2 months)

Developed the company's first machine learning platform for multi-objective physics-based optimization of SSD, resulting in a 10% reduction in design costs. Led a cross-functional team to integrate machine learning automation into SSD integration, reducing iteration time by15%. Deployed Convolutional Neural Network (CNN) algorithms for handwriting recognition in the manufacturing process, automating data processing with a 95% increase in efficiency, which resulted in a 40% reduction in manual processing costs and improved accuracy by 98%. Applied physics-informed neural networks to predict semiconductor package failure probabilities, reducing customer testing failure rates by 20% and cutting testing-related costs by 30%.

Machine Learning Engineer Intern

Western Digital

May 2022 - Aug 2022 (4 months)

Designed feature extraction pipelines using machine learning on Linux, leading to a 30% reduction in manual data processing time of identifying key SSD design features. Implemented deep learning models in TensorFlow to predict SSD failures, increasing prediction speed by 10000x compared to traditional numerical methods. Developed and validated machine learning models to simulate NAND warpage with 93% accuracy.

C

Research Assistant

Computation Material Lab (University of Wyoming)

Sep 2020 - May 2023 (2 years 9 months)

Implemented parallelized Generalized Finite Element Method (GFEM) algorithms in C++ with Open MPI, scaling simulations to handle 10 million data points with a 90% reduction in computation time. Developed reduced-order models to accelerate simulations by 1000x, facilitating rapid material design and optimization. Optimized 3D nonlinear materials using advanced GFEM and gradient-based optimization techniques.

A

Research Assistant

Advanced Ultrasonic Imaging Lab (University of Wyoming)

Sep 2018 - Aug 2020 (2 years)

Engineered physics-inspired machine learning models in PyTorch for predicting brain blood clots from ultrasonic time-series data, achieving 90% accuracy. Developed fast inversion tomography algorithms for corrosion detection using Python and GPU acceleration, reducing computation time by 100x. Applied CNNs and transfer learning for wavefield (video) reconstruction from sparse sensor data, optimizing signal processing efficiency.

About

Detail

Timeline

Résumé