I have worked in multiple fields including quantitative analysis, signal processing, and software development, etc.
I am always passionate about new ideas, innovations and technologies, and keep expanding my interests
and growing my professional skills.
Areas of Interest
- Quantitative research and algorithm development
- Machine learning and data analytics
- Software development
- Quantitative analysis, data analytics, algorithm development, numerical simulation and back-testing.
- Python scientific/data stack: pandas, numpy, statsmodels, jupyter, etc.
- Machine learning algorithms and software libraries: scikit-learn, xgboost, keras, etc.
- Python software development with object-oriented design. Experienced in SQL, C/C++, Matlab, R, Linux.
- Applied statistics, estimation, inference and signal processing.
- Quantitative Research Analyst; Quantitative Software Engineer, iFDC, Irvine, California, 2017-2020.
- Researched on various machine learning models, including random forest, gradient boosting decision and support-vector machine (SVM), etc., to classify market trending patterns and reduce oscillating loss to optimize momentum strategies.
- Developed and back-tested mixed-timeframe multi-asset commercial portfolio management strategies, based on time-series momentum or mean-reversion models.
- Analyzed hundreds of GBs of time series data to build historical volatility distributions, ticker correlation distributions, and extract effective features to improve transaction models.
- Developed adaptive execution algorithms for time-sensitive execution to minimize timing risk.
- Implemented various strategies and algorithms as fully automatic execution modules in Python with event-driven micro-service architecture and deployed them in production on cloud platforms.
- Developed trading desk tools in Python for real-time market data statistics, indicator signaling, historical database access, market data API streaming and user front-end visualization.
- Staff Engineer - Systems; Sr. Staff Scientist - Software Systems, Broadcom, Irvine, California, 2010-2017.
- Renovated the statistical model and signal processing algorithm for higher accuracy of the dynamic temperature-compensating GPS receiver clock using Kalman filtering and polynomial regression, based on analytics and modeling of large data sets from lab experiments.
- Developed various numerical programs in Python to optimize 3G/LTE mobile radio chips, e.g., fitting filter coefficients with lab data regression, searching circuit non-linearity compensating coefficients, estimating signal-to-noise ratio from signal samples, adjusting feedback delay mismatch of closed-loop power control, etc.
- Developed GPS receiver host software in C++ and signal processing firmware in C to enable and calibrate GPS radio modules.
- Log analysis and processing with Python, and software/algorithm defect tracking via JIRA for GPS host software and signal processing algorithm firmware.
- Developed a comprehensive set of automation tools in Python for RFIC tests. The tools manage testing case, error recovery, instruments control, and data report & visualization.
- Algorithm Engineer, Datang Mobile, Beijing, China, 2003-2006.
- Innovated on wireless channel parameter estimation and signal detection algorithms, which significantly suppressed interference and improved system capacity (U.S. patent 8023486 B2).
- Developed a complete set of numerical simulation programs with C/C++/MATLAB for evaluating and optimizing 3G TD-SCDMA modem algorithms and models.
- Ph.D. Electrical Engineering, University of California Riverside. December, 2010.
Dissertation: Performance of time delay estimation and range-based localization in wireless channels. PDF
- Researched on low-complexity wireless signal source geo-location algorithm, and biased statistical models and estimation methods for wireless ranging.
- Found new theoretical bounds on the performance of wireless time-delay estimation in unknown random multipath channel with PPM and frequency-hopping signals.
- Innovated on highly efficient numerical algorithm to compute time-delay estimation bounds, based on the moment-generating function of quadratic Gaussian distribution. The Matlab code exploiting special matrix structures speeds up by more than ten times than regular methods.
- M.E. Information & Communications Engineering, Xidian University, Xi'an, China.
- B.E. Telecommunications Engineering, Xidian University, Xi'an, China.