Project: Optimizing Airline Crew Scheduling for #11 airline company in the U.S | Python, Azure Databricks, PowerBI• Engineered inefficiency (excessive layover and deadhead time) detection program in Databricks from high-dimensional data, designed real-time, interactive dashboard that empowers swift decision-makingProject: Customer Segmentation Strategies for Non-Profit and Community Development Financial Institutions | Python, R, Tableau• Identified key characteristics in customer segmentation using clustering technique for better customer target, mainly for low-income customer to enhance Homeownership (15% increase), FICO Credit Scores (130-point rise), and Savings (127% increase).Project: Impact Analysis for College and Career Readiness Program | Python, R, Tableau• Developed XGBoost and 5+ machine learning models to discover crucial factors effecting students' post-high school successes with accuracy of 97%, designed interactive Tableau dashboard, enabling real-time of curriculum modification decisions.

𝗩𝗣𝗕𝗔𝗡𝗞 - 𝗦𝗨𝗠𝗜𝗧𝗢𝗠𝗢 𝗠𝗜𝗧𝗦𝗨𝗜 𝗕𝗔𝗡𝗞𝗜𝗡𝗚 𝗖𝗢𝗥𝗣𝗢𝗥𝗔𝗧𝗜𝗢𝗡#1 consumer lending, personal loans & consumer fintech market leader in Vietnam

Main Responsibilities- Monitor and analyze portfolio quality by risk indicators, net profit margin and make budget planning to ensure the optimal portfolio structure, prepare monthly reports with insights and proposals to Risk Management Committee. - Perform segmental analysis by credit score, behavior score, demographics, application info, employment type by social insurance status, credit history by credit bureau,… in order to figure out the root causes and identify the loss rate correlation with product features, credit policy to propose appropriate underwriting process.Results & Recognition- Finish the NPM analysis by products, decide the cut-off by applying new score range for each segments achieve estimated 20% risk cost decrease in line with estimated 15% net profit increase.- Design a implement successfully framework analysis of social insurance with multiple dimensions: government/private company or self-employ,.. have job/job loss status, company size for each customer to match with risk rank - Build income model to estimate actual income of customer by combing social insurance amount and income tax amount of customer. Currently cover 30% total company with size 500+ over Vietnam with variety of application such as Debt burden ratio calculation, qualified lead generation , underwriting process, credit limit increase,…

- Predue scorecard model for credit card – predict propability of not repayment in due, in order to classify good/bad customer to no-call/call strategy in line with proper strategy, final GINI of the model is 74% - higher 5% than previous model in both training and validation period, improve the Roll-rate to next bucket around 0.7% with same capacity- Debt burden ratio (DBR) based on social insurance – structure and combine various kind of data from vendor (credit bureau, social insurance provider,…) to calculate current debt of all loans over the income of customer, crossing between DBR and behavior score to calculate the maximum increased limit for customer. The results of expected loss for increase limit based on DBR is about 1.7% for 44,000 accounts with increased limit range from 10 mil – 25mil VND.- Social insurance check – mining, analyzing to make the vintage analysis, performance comparison between job loss, job having and not have social insurance record segment. Applying successfully and take action in time especially after Covid effect, prevent the loss rate coming from loss job segment which is almost 2 times compared with remaining segment.- Propensity model for Credit Card Cross Sale – Top up based machine learning model – XGBOOST increasing conversion rate by 0.15% and saving resource. Final GINI of the model is 65% on train sample and 63% on validation sample.- R script for a loop to facilitate effectively PSI (Population Stability Index) calculation reducing repeated steps andauto-generate data frame according to given data set to achieve final PSI results in less than 3 minutes with around 5 million records.- Predictive model of underwriting process for New to bank customer, the number of training observations is 1.3 million with traditional approach – Logistic regression gaining the GINI 37%- Predue scorecard model for collections and improve the discriminatory power (GINI) of model by 10% after 3 years without revision

Main Responsibilities- Work with IT for integration of auto-calling system from vendor and current calling system, initiate the call flow fullfil both business requirement and technical ability in Voice Virtual Assistant and Voice Analytics project- Key PIC for UAT data preparation, implementation, change request in Pega collection project, balance between business requirement and system ability and convince involved party with logic and optimal process in all projects I involved. Results & Recognition- Voice virtual assistant (Go-live) one of the highest priority project to save operating and provision cost up to 1,9 bil VND per year and reduce humance intervention. - Voice analytics (Go-live), increase agent collection efficiency 0.25% after report insight from project and improve customer experience after sales service as well.- Replace sucessfully old system with new Pega system with over advantages such as auto-allocation to caller, auto-generated trail reports, easy modification of business rule instead of fixed code in old system.

• Involved in auditing the sections (Cash, Plant Property Equipment, Account Receivable/Payable,..) in financial statement of FDI companies, largest pharmaceutical company in Vietnam by following the auditing procedure. Ensure that all the figures in the financial statement of client are reliable and reasonable after auditing. • Involved in final recommendation to the clients by consolidating flaws of recording the expense, (duplicating, misclassification, etc.) for client to improve cost efficiency to optimize regular internal reviews or audits to maintain continuous compliance and readiness.