Mingming Liu Email & Phone Number

California, United States

San Jose, CA, US

• Developed a probability of default model (XGboost) achieving a 99% capture rate. Integrated this model into line adjustment strategies and loss forecasting, resulting in over $15M loss reduction and $140M incremental.
• Led the design of collection-related KPIs (outreach, RPC, PTP) for the leadership team, automating reporting and visualization. Analyzed the most collectible portfolios for strategy design and validated collections.
• Developed and visualized credit and fraud risk KPIs using SQL and Tableau, enhancing operational efficiency by over 50%. Regularly tracked potential risk exposures and implemented proactive strategies, cutting.
• Key contributor to a credit card preapproval campaign, driving $6M in annual revenue. Focused on new customer acquisition and accounts default risk analysis, collaborating with sales and engineering teams to enhance.
• Mentored a team in migrating dashboards

San Jose, CA, US

Design Credit and Fraud KPI, Automate Reporting, Build PD model and strategies

Palo Alto, California, US

Toronto, Ontario, CA

• Led credit card late fee and credit risk research as project leader. Conducted segmentation analysis for accounts, assigning them different tiers, and developed loss forecasting models based on tiers, increasing.
• Forecasted deposits using linear regression and moving-average models, tested from statistical (cross-validation) and financial perspectives. Implemented model using SAS, achieving prediction error under 10%.
• Participated in balance sheet forecasting for stress testing, conducting ad-hoc analysis and applying overlays to forecasting results, and presenting analysis to the business department.
• Coordinated with the validation team to monitor 10+ models performance annually(OOT backtesting, statistical testing)

Wayne, Pennsylvania, US

• Developed AR and stochastic model to forecast economic indicators, implemented gradient descent in C++ to get parameters, controlling relative MSE less than 5 bps, validated model by back-testing & benchmarking
• Built delinquency model by logit & probit & cloglog regression, obtained accuracy over 70%, compared with NN & SVM & Random Forest models to analyze performance, combined with Markov Chain to generate cash flow and.
• Conducted Monte Carlo Simulation on AWS, generated risk reports and cooperated with pricing team in pricing products
• Develop pricing model to evaluate MBS and implement in python, deliver purposeful analysis promptly

New York, New York, US

New York, New York, US

Master Of Arts - Ma, Statistics

Bachelor Of Science - Bs, Financial Mathematics

Exchange Student, Statistics