Edson Leung

Built data processing pipelines and optimized retrieval strategies (chunking, metadata filtering, top-k optimization) for a production RAG system, improving retrieval relevance scores by 10-20%. Developed an evaluation framework using RAG metrics to identify low-faithfulness patterns, improving answer quality scores by 0.1-0.2. Supported production deployment by validating data pipelines and catching integration issues (dependency errors, prompt template bugs, memory limits), reducing runtime failures and rollback incidents.

Built and optimized ETL pipelines with Apache Airflow and BigQuery, reducing data processing time by 40% and automating data flow from Salesforce CRM, APIs, and Google Analytics. Created scalable data lake architecture on Google Cloud to store SKU, production, and logistics data, enabling flexible downstream reporting and model development. Built real-time KPI dashboards on Looker for operations, marketing, and finance, saving over 20 hours per week of manual reporting.

Conducted comprehensive analysis of customer purchasing patterns, sales trends, and funnel metrics using SQL, driving a 15% sales increase by optimizing personalization strategies for a customer base of 4M+. Enhanced customer segmentation and targeted campaigns through K-means clustering and RFM analysis with Python, boosting campaign engagement by 35%. Designed a churn prediction model to proactively identify at-risk customers, enabling retention strategies that improved retention rates by ~20%.