Lihua L

I do things to improve user experience at Twitter Home.

1. Lead the DS effort on multiple product teams and biz ops team: achieved a 8% MoM in user profile and preference data collection for member experience team; had improved retention by 43%YoY, define KPIs, develop tracking framework and evaluate potential tradeoffs within the next generation of personalized recruiting and talent matching product ("GD Wallet") for mobile app team; increased total engagement apply starts by 8% and job visits by 10% QoQ; complete build and validation of data sources for overall retention dashboard and consumer traffic forecasting model. 2. Participate in the product and engineering planning, partner with PMs to define the roadmap for product and data science team. 3. Lead the effort to communicate state of the business to stakeholders regularly, and provide insights to drive strategic decisions, evaluate and define key product and business metrics. 4. Build key data sets/pipeline to empower operational and exploratory analysis, build and automate dashboards and reports for executives and business units. 5. Design/setup AB tests and guide product decisions based on data and testing results. 6. Manage a team of Data Scientists, and currently lead DS effort in multiple business functions (Mobile, Engagement, Content ML, Job Search ML and Knowledge Graph).

1. Convert DFAST/CCAR models to fulfill CECL standards. 2. Develop predictive models that estimate probability of default (PD), loss given default (LGD), and exposure at default (EAD) for each residential mortgage loan and home equity line of credit on the bank’s balance sheet. Models gets highest possible rating from Model Validation team and are well received by FDIC. 3. Develop and maintain internal R packages, which house all custom functions for model development, transition matrix, cash flow engine, back-testing, model validation, statistical tests and sensitivity analyses. 4. Translate business problems into statistical problems, and develop models to support business needs, such as pricing policy, cross-sell targeting, credit risk modeling, scorecards. 5. Perform ad-hoc analysis including business intelligence on customer behaviors, general business support through data collection and processing, summarize insights, create dashboard and present key findings to senior management. 6. Manage a team of predictive modelers, lead meetings and discussions with senior management, Model Risk Management, and bank regulators about model purpose, data preparation, statistical methodologies, model results and internal controls.

1. Took initiation to develop a new solution named "COMET" (Constrained Optimizing Model Estimation Technique) to build scorecard models using R and Shiny, whose performance is comparable to Xeno. The technique is later upgraded from a research project to a standard product offering of TransUnion, and now is being used full-time. 2. Applied analytical skills to work on all aspects of the account lifecycle in the consumer credit domain on behalf of a diverse set of clients, ranging from marketing and propensity models for customer acquisition and retention, credit risk models for acquisition and account management, cross-sell applications, portfolio models for regulatory applications, event-based trigger solutions, and strategy analyses of various kinds. 3. Performed extensive work with large dataset, explored new data sources and statistical and machine learning methodologies; 4. Developed functions and packages in R, collaborated with other developers using Git.

1. Built demos utilizing SAP Predictive Analysis, coded proprietary SQL and R-Programs for SAP HANA-based planning scenarios and presented to the whole company weekly; 2. Worked on a lemonade sales project and optimized the daily inventory planning if running a promotion using real-time method based on the Amazon Web Services Cloud. Work is presented at the Chicago ASUG meeting and would be posted on Amazon to show the advantage of dealing with bigdata of HANA to the market; 3. Trained the company on the SAP HANA tools after research.