David Sung

Programming Languages: Python (Pandas, PySpark, Numpy), SQL
Data Processing Frameworks: Apache Spark, Databricks
Machine Learning: XGBoost, TensorFlow, scikit-learn, Linear regression/Logistic Regression, Clustering, Decision Tree
ETL Tools: AWS Glue, AWS Data Pipeline
Data Storage: AWS Redshift, MySQL, PostgreSQL, Oracle, Hadoop Distributed File System (HDFS)
Monitoring Tools: Prometheus, Grafana, Elastic Stack
Big Data Tools: Apache Airflow, Apache Hive, Presto, Apache Hadoop, Apache Kafka, Apache Zookeeper, Gobblin, Apache Sqoop
Data Visualization Tools: Tableau, Power BI, Looker
Statistical Analysis: Hypothesis Testing, Regression Analysis, ANOVA, Probability Theory, Conditional Probability, Probability Distributions, Causal Inference, Statistical Data Analysis, Data Mining, Predictive Modeling, Time Series Analysis
Analytical Expertise: A/B Testing, Experimental Design, Qualitative/Quantitative Analytics, Data-Driven Decision Making, Strategic Planning, Data Visualization, User Behavior Analysis

Led the enhancement of Mastercard's Smart Authentication model within the EMV 3-D Secure system, analyzing data trends to guide project direction, improving security and UX through innovative technology applications.
Led a cross-functional team of 15 professionals including Product Managers, Data Scientists and Data Engineers through the entire Product Development Life Cycle by planning, managing risks, and enforcing quality standards.
Addressed regional model variations through data trends and statistical analysis, leveraging Apache Hive, Pandas, and statistical analysis techniques like hypothesis testing and predictive modeling.
This resulted in a 95% fraud detection rate, a 20% increase in global customer satisfaction, and an estimated $10M annually in fraud prevention.

Developed next-generation data-monitoring tools to track the health status of various Hadoop clusters and detect data node volume failures.
Employed Apache Kafka for real-time data streaming and for analyzing these streams to pinpoint anomalies, and Apache Gobblin to integrate and manage data sources across the infrastructure, with an uptime of 99.95% for critical services.
Engineered data pipelines using Apache Airflow for workflow orchestration and AWS Data Pipeline for efficient data transfer across AWS services, ensuring data consistency and achieving a peak response time of under 100 ms.
Optimized data infrastructure operations by automating routine tasks, cleaning out storage space, and organizing tables and schemas on MySQL servers in preparation for migrating Spark/Hive jobs to the cloud, thereby reducing manual maintenance time by 15 hours weekly.
Implemented ETL processes using AWS Glue to efficiently and reliably move data across our cloud infrastructure, with a data throughput of 250 GB/hour from operational databases to the data lake for real-time analytics.
Improved database performance and streamlined data pipeline operations by integrating Apache Spark for rapid data processing, using Presto for low-latency data querying across heterogeneous sources, and optimizing AWS Redshift for efficient large-scale data warehousing, reducing query execution times by 40%.
Developed a comprehensive monitoring system using Prometheus for collecting metrics, Grafana for visualization, and Elastic Stack for log processing, to continuously track the health status of Hadoop clusters and data pipeline tools.
This integrated approach achieved an impressive average issue detection time of under 5 minutes.

Utilized Python and data tools such as Pandas and NumPy to scale memory margining processes across approximately 12.9GB of data, enhancing the performance of both 1-socket and 2-socket Open Compute Project (OCP) server designs in Facebook's data centers.
Employed A/B testing to optimize configurations, impacting billions of users and optimizing data throughput by 15%.
Crafted high-impact visualizations in Tableau to convey the overall quality metrics of the server design, utilizing Hive SQL-like queries to extract and analyze data.
Integrated results from A/B testing experiments to demonstrate comparative effectiveness of server configurations, enhancing decision-making processes.
Achieved Facebook’s OCP debug card validation within 3 weeks by defining a test plan using hardware testing tools like Oscilloscopes and Logic Analyzers, and forming a strategic alliance with Facebook’s hardware engineers and external ODM vendors to drive the project to completion.

Led the Advantech Data Center Networking and Communication Systems Launch by deploying AdvancedTCA-ATCA, MicroTCA, CPCI, and AdvancedMC–AMC technologies, achieving key performance targets and significantly enhancing client satisfaction through improved system reliability and performance.
Developed a comprehensive project management plan, defining KPIs for scope, schedule, and cost, which directly influenced project tracking and resource allocation.
Established a strategic communication plan to maintain continuous engagement and clear guidance across all project phases.
Enhanced British Telecom's system performance by 80% in 2016 through firmware enhancement leveraging Intel DPDK, which optimized network packet processing capabilities, thereby saving a $25 million contract and solidifying a strategic partnership.

Led the Smart Authentication Airflow Project, driving data-driven decision-making in product development by implementing automated data reporting processes that streamlined data workflows.
Managed over 30 diverse Decision Intelligence (DI) model work requests, focusing on translating complex data into strategic decisions for product optimization and enhancing customer satisfaction.
Liaised closely with Marketing, Sales, and Engineering teams to translate data insights into actionable strategies that align cross-departmental goals with product objectives.
Leveraged advanced statistical tools and techniques, such as Python and SQL for data manipulation, enhanced by A/B testing and user behavior analysis, to deeply understand credit/debit card transaction patterns.
This comprehensive analysis not only led to actionable insights but also drove substantial business growth by improving user engagement, optimizing conversion rates, and informing pivotal strategic product decisions.
Designed and interpreted A/B tests using Google Optimize and Optimizely, focusing on user behavior analysis to enhance the understanding of user interactions and preferences.
This validated product hypotheses, leading to targeted product improvements that impacted user engagement and retention in card authentication and authorization models.
Analyzed user behavior data with tools like Mixpanel and Amplitude to highlight key user actions and trends using methodologies such as cohort analysis and churn prediction.
This directly influenced product strategies that significantly boosted user engagement and retention in our card authentication and authorization systems.
Presented proof of concepts to customers using blind data from production servers to showcase model efficacy in fraud detection, effectively translating complex data findings into understandable visual representations and rebuilding fraud detection models with new features upon significant changes in customer spending behavior.
Developed and monitored key performance indicators for data initiatives using Prometheus, Grafana, and the Elastic Stack, and visualized through Tableau, Power BI, and Looker, resulting in comprehensive monthly reports highlighting fraud prevention savings, authentication friction, and decline rates for executive management.
Oversaw the implementation of data governance policies and compliance with data protection regulations, ensuring the security and privacy of sensitive data across all platforms.
Actively engaged in continuous professional development by attending workshops on Advanced Machine Learning and Predictive Analytics, pursuing certifications such as Certified Analytics Professional (CAP), exploring cutting-edge data science techniques like deep learning and ensemble methods to bring innovative solutions back to the team.
Processed large volumes of transactional and authorization data using scalable data solutions to ensure secure storage, accessibility, and data integrity.
Developed automation scripts with Python and SQL for anomaly detection utilizing moving averages and standard deviation analysis to clean and preprocess data by calculating the 7-day moving average and identifying outliers when score distribution was outside of 1.5 standard deviations, ensuring data quality and integrity for monitoring smart authentication score distribution across various regions.
Generated actionable reports that identified root causes of score discrepancies, leading to a 20% reduction in transaction discrepancies over 6 months.
Developed machine learning models using XGBoost and Python, enhancing fraud detection accuracy in Mastercard’s Decision Intelligence Model for all card transactions to 92% and reducing false positives by 15%, utilizing transaction data, fraud detection data, cardholder data, device data, and merchant data sourced from fraud detection systems.