I am a principal data scientist at AT&T. Before joining AT&T, I was a staff research engineer and manager at Samsung Research America, developing signal processing/machine learning algorithms for smart-home, IoT, smart devices, and wearable healthcare projects. 

I received the Ph.D. degree in Electrical Engineering (with focus on signal processing and machine learning) from University of California, San Diego (UCSD) in 2012.  I am also an elected committee member of IEEE Bio-Imaging and Signal Processing Technical Committee, an associate editor of IEEE Journal of Translational Engineering in Health and Medicine, and an IEEE Senior Member. Since 2015, I have been an adjunct professor with School of Computer Science and Engineering, University of Electronic Science and Technology of China.

I have strong interest in Big Data for finance, Big Data for advertising/marketing, and Big Data for healthcare. My expertise includes (some old research can be found at here) 
  • Signal Processing
    • Sparse Signal Reconstruction/Compressed Sensing
    • Signal Decomposition, Separation, and Representation
    • Statistical and Adaptive Signal Processing
    • Noise Removal
  • Machine Learning
    • Sparse Bayesian Learning
    • High Dimensional Regression
    • Feature Selection and Extraction
    • Artificial Neural Network and Deep Learning
  • Time Series Analysis
    • Dynamic and Spatio-Temporal Modeling
    • Statistical Forecasting
    • Motif Discovery
  • Quantitative Finance
    • Signal Processing and Artificial Intelligence for Algorithmic Trading
    • Risk Management
    • Modern Portfolio Theory
My research/development work was always driven by challenges in machine learning and signal processing. In past years I focused on the following challenges with successful solutions:
  • Challenge in Machine Learning: Sparse Bayesian Learning for Compressed Sensing and High-Dimensional Regression
    • Proposed a sparse Bayesian learning framework which explores and exploits correlation structures in the underlying solutions of compressed sensing models. The framework greatly relaxes the traditional requirement in compressed sensing that underlying solutions must be sparse or sparse in transformed domains. Based on this framework, more practical complex and high-dimensional regression problems can be solved with tolerated errors. Published articles were cited over 1500 times. Two publications on its applications to compressed sensing of noisy biosignals were ranked as the "Most Cited Articles during 2013-2014" on top IEEE journals.