A graduate level course project of Algorithms and Data Models developed in a team of 3 in University of Victoria
Analyzed data by using relational data model, conceptual design, normalization, graph data model, and triangulation to conclude the insight found from the data by using Python and SQL
A graduate level course project of Algorithms and Data Models developed in a team of 3 in University of Victoria
Discovered four types of biases and four limitations by using Python Natural Language Toolkit (NLTK), SQL, word embedding, and clustering in Twitter Discourse
A graduate level individual project of Optimization for Machine Learning course in University of Victoria
Compared the performance of optimization algorithms in different cost functions, in different optimization algorithms, and in different features of input data
A graduate level individual project of Deep Reinforcement Learning course at University of Victoria
Improved up to 4.7% of maximum drawdown to maximize the profit in the stock market of the Vanguard 500 Index Fund ETF (NYSE: VOO) by using the market information, technical indicators, economic indicators, and Deep Deterministic Policy Gradient method
Co-designed the world’s first AI-SDR-to-HDR algorithms implementing on flagship SoC Dimensity series for mobile devices (customers: Xiaomi, OPPO, Vivo, HONOR)
Solved stability challenges in AI-SDR-to-HDR algorithms by applying novel approaches
Implemented on MediaTek flagship Dimensity series [Hyperlink1] [Hyperlink2]
Formulated algorithms for the next generation HDR video technology and dynamic contrast algorithms, implementing SoC Helio P series and Dimensity series for mobile devices
Increased deep learning model accuracy by 30% and reduced labeling time by 90% for benchmarking noise, texture, brightness, and contrast by using image processing algorithms
An M.S. thesis in computer vision applications
Conducted camera calibration, multi-camera image rectification, and view synthesis algorithms
Published in an international conference [Hyperlink]
A personal project applying image processing algorithm and hardware design
Designed novel hardware-oriented color de-mosaicking algorithms, saving at least 11.2 % gate counts and 91.5 % power consumption, and improving the average CPSNR by over 1.785 dB
Commercialized into an intellectual property (IP) by a semiconductor IP development company [Hyperlink]
Published in the IEEE TCASII journal [Hyperlink]