• Drive the research agenda and lead a team of data scientists to conduct research using state of art AI methods (DL and traditional ML) to build next generation search solutions.• Demo AI product to technical leaders and C-level executives at different companies. • Lead consulting projects including resource management, SOW planning, advisory or hands on implementation of ML solutions. • Help enable ML training and adaptation through out the company. Lead the effort of data scientist recruiting.

• Designed and built automatic document clustering, anomaly detection, topic labeling and NLP modules (e.g., spell checker, query rewriting, synonym detection) for Lucidworks Fusion AI product using Spark, Scala and python.• Lead consulting projects to help customers improve search relevancy using ML approaches such as query analytics, recommender and clickstream models.

Involved in software development, POC consulting, testing, documentation and training of SAS Model Implementation Platform, which is an in-memory multi-threaded analytic engine that enables fast model implementation calculations on a gridded network.

Initiated and lead the "Robot Modeler" project, with a goal to build an automatic model-building module to reduce the modeling time from months to days.Experimented different methods to build Chase 360 index, the Chase XLOB credit score.Served as committee for PHD technical interview events.

Text Mining: 1. Developed string comparison algorithms for transaction data mining and job changer identification; Reduced implementation time from 1000+ hrs to 1 hr by Perl programming and parallel computing. The analysis has been put into production database for firm wide use. 2. Sentiment analysis research for text data from customer surveys, social media and acct memo; Improved hit rate by 30% compared to SAS text miner and provided automatic keywords extraction and weighting.Predictive Modeling: 1. CRM: Support Chase Merchandise Service marketing by predicting customer spending behavior using Chase credit card and mortgage data. 2. Fraud Detection: Built Tier 1 model to detect check fraud activities, which results in $80 million annual value saving; Helped with DDA collection strategy model. Risk Management (CCAR): 1. Built Chase Credit Card Risk Weighted Asset forecast model; 2. Help Build and calibrate Chase mortgage risk PD model.

 Build statistical models and software to visualize high dimensional data. Create an interactive visualization for NSF award portfolios.Demo:http://www.youtube.com/watch?v=n5rNV5Q7O5ohttp://www.youtube.com/watch?v=ud-jfB87hDA Conducted large scale data analysis of 100 years precipitation records in East Africa. Modeled inter-events data using stochastic process approach.

 Explored different clustering methods for large cellular network data (475,000 users). Explained the performance of clustering algorithms using simulation study.

 Increased calculation efficiency by developing a Streamline Immuno Assay Cutoff Determination Package. Designed GSK XLfit software training material.

 Assisted Virginia Tech researchers with data analysis, results interpretation, software coding, report writing, and proposal preparation.