DeepToF: Neural network development, evaluation and fine-tuning for depth estimation based on ToF raw data.• Investigations on state-of-the-art NN architectures to come up with an NN architecture robust in depth accuracy and light in computing power consumption, for eventual deployment of the algorithms on ToF cameras of mobile devices.• Pre-training and fine-tuning of various NN architectures to evaluate their performance both in overall MAE and in tackling key problems in ToF… Show more DeepToF: Neural network development, evaluation and fine-tuning for depth estimation based on ToF raw data.• Investigations on state-of-the-art NN architectures to come up with an NN architecture robust in depth accuracy and light in computing power consumption, for eventual deployment of the algorithms on ToF cameras of mobile devices.• Pre-training and fine-tuning of various NN architectures to evaluate their performance both in overall MAE and in tackling key problems in ToF depth estimation, including denoising and MPI.• Attempts of developing a novel NN architecture aggregating the advantages of various state-of-the-art architectures to achieve a better balance in depth accuracy and computing power.• Looking into alternative solutions, particularly traditional solutions to the MPI issue. Show less

Extracurricular Experiences:• Coursework: Machine Learning (Andrew Ng, Coursera); Deep Learning Specialization (Andrew Ng, Coursera); Kaggle Learning Micro-courses: Python, Pandas, Machine Learning (scikit-learn), Deep Learning (tensorflow), Data Visualization, SQL;• Completed two Kaggle machine learning projects by developing robust supervised classification and regression machine learning models to predict the survivability of Passengers on Titanic and the housing prices in Ames… Show more Extracurricular Experiences:• Coursework: Machine Learning (Andrew Ng, Coursera); Deep Learning Specialization (Andrew Ng, Coursera); Kaggle Learning Micro-courses: Python, Pandas, Machine Learning (scikit-learn), Deep Learning (tensorflow), Data Visualization, SQL;• Completed two Kaggle machine learning projects by developing robust supervised classification and regression machine learning models to predict the survivability of Passengers on Titanic and the housing prices in Ames, IA.• Conquered two Kaggle deep learning projects on computer vision, by developing robust convolution neural network (CNN) models to recognize hand-written digit patterns and identify metastatic cancer from digital pathology scans.PhD Research Experiences:• Formulated and validated a novel prognostic model integrating machine-learning-based anomaly detection and Bayesian-inference-based predictive modeling, for condition-based maintenance on output degradations of IGBT modules and of RF GaN HEMTs under accelerated life tests (ALTs).• Conducted an unprecedented investigation on ohmic contact diffusion of GaN HEMTs under RF-ALTs including GaN HEMT device modeling and failure mode simulation, to model the complete diffusion process and behavior.• Investigated for the first time the radiation hardness of commercial E-mode GaN FETs under various doses of gamma radiation to develop the next-generation devices for high-power RF and microwave applications.• Developed a novel and expandable Bayesian estimation model for preliminary end-of-life (EOL) prediction of power electronic devices.• Analyzed and compared the system reliability of offshore wind turbines with five different drive train and electrical system configuration using reliability data from handbooks (MIL-HDBK-217 and OREDA), FMECA and fault tree analysis. Show less

• Delivered lectures on Bayesian estimation and PhD research on prognistics for graduate-level failure mechanism course ENRE600 and its undergraduate equivalent ENMA426• Organized and lectured the review sessions for undergraduate-level material science and engineering course ENME382

• Re-developed and optimized an automated data generator, collecting metrology and 3D data from an etch simulation and performs finite-difference time-domain (FDTD) sweep simulation to generate spectral data.• Completed re-development of the python-based workflow of the data generator/analyzer to enable parallel processing of batch jobs of etching multiple wafers simultaneously.• Set up and validated the parallel processing of the data generator on both a single PC and up to 10… Show more • Re-developed and optimized an automated data generator, collecting metrology and 3D data from an etch simulation and performs finite-difference time-domain (FDTD) sweep simulation to generate spectral data.• Completed re-development of the python-based workflow of the data generator/analyzer to enable parallel processing of batch jobs of etching multiple wafers simultaneously.• Set up and validated the parallel processing of the data generator on both a single PC and up to 10 virtual machines on Microsoft Azure platform, recording an improvement of up to over 3× faster in computation time.• Modified the process flow to generate STL mesh file instead of binary 3D files, and validated the workflow using binary 3D data for FDTD simulation by recording over 40% improvement in speed over the one using STL files. Show less

• Specialized in the design improvement process of a new office desk product series of Haworth• Completed improved 2-D engineering drawings in AutoCAD and 3-D CAD model of the new office desk product in Solidworks