Lead a team of 20+ data scientists and data engineers to apply AI and machine learning to drive key business initiatives and help Adobe make data-driven decisions.

Online advertising R&D, large-scale behavior targeting, Machine learning algorithms* Led the improvement of the predictive segment platform which performs user level prediction and optimization based on AOL’s rich user behavior data (millions of users and thousands of features)* Prototyped and developed large-scale machine learning algorithms to to create custom user segments for ad campaigns to target, which have 4-8 times higher conversion rates than non-targeted user group.* Conducted various ad hoc analysis projects using MapReduce on tera-bytes of user browsing history to capture user behavior, network activity, purchase intention and so on.* Built a monitoring/reporting framework for predictive segments, providing weekly and monthly performance reports to management

Portfolio risk modeling* Developed portfolio risk score to predict the risk of accounts in short term/long term using data mining techniques and logistic regression* Built optimal credit line model to maximize the profitability of accounts and provide guidance for line increase/decrease decision using SAS/OR, 15% increase on the ROI showed by initial study.

Calibration of profit forecasting model (PFM) * Utilized the Multinomial Logit Model to model customers’ choices among different itineraries in the airline markets. * Obtained Maximum Likelihood estimators of parameters using large amount of historical data by programming in GAUSS. * Introduced market share constraints to the maximum likelihood estimation and improved the model performance by 20% in terms of profit prediction.

My research interest includes Stochastic Modeling and Optimization, Reliability and Closed-Loop Supply Chain. As a research assistant, I worked with Prof. Sarah M. Ryan on a NSF project, "Auto-steered Information-Decision Processes for Electric System Asset Management", in collaboration with researchers from electrical, computer engineering, statistics and computer science at Iowa State University. The following model can help the supplier in a product-service system make good joint decision on asset management and inventory control.-- Product-service system* Formulated a stochastic model to investigate the joint optimization problem of asset management and inventory control in the context of service-oriented business model* Developed and implemented optimization methods and heuristics. Experienced with numerical optimization (Newton’s method, quasi-Newton method, BFGS) and nonlinear optimization * Evaluated the effect of categorizing the returned products in remanufacturing and showed that quality-based categorizing can reduce costThe following model and methodology can assist engineers and managers to assess the potential value gained from condition monitoring and to make better decisions on the management of high-voltage transformers as well as other valuable assets.-- Condition-based maintenance* Developed and optimized condition-based (data-driven) maintenance policies for large power transformers * Experienced with several statistical models: proportional hazards model, continuous time Markov chain.* Deducted complicated mathematic formula for the policy * Assessed the value of condition monitoring in power industry and conducted sensitivity analysis to identify the most influential parameters on the costKey Skills Stochastic Modeling and System Analysis, Optimization, Reliability, Production Planning and Inventory Control, Monte Carlo Simulation

Research focuses on visualization of large datasets and airline database analysis.Airline database analysis* Exploited patterns, trends and questions using R and SQL* Predicted plane delays using weather conditions* Determined the best time to fly to minimize delaysVisualization of large-scale database* Built interface between R and large-scale database to visualize the variables in the database* Determined the proper data resolutions to render a good graphic* Implemented R modules to automatically scale down entries in the database and produce various statistical graphics Key Skills Data Visualization, R, SQL, Sufficient Statistics

Research focuses on computer vision, artificial intelligence and electronic design. Face detection* Developed a face detection framework using haar-feature based face detection algorithm* Built up a three-phase multi-view face detector by introducing a pose estimator (using fisher linear discriminant)Waveform regenerator* Conceptualized, designed and fabricated a waveform regenerator for a medical instrument (simulate the signal generated by hand pressing on a metal board)* Developed and tested the hardware and software for data acquiring, data processing and data regeneration.Key SkillsImage processing, Pattern recognition, C++, C#, Assembly language, Micro-controller programming, Labview 7.0 (National instruments)