• Developed an anomaly detection machine learning model for container services to proactively discern customers undergoing migration to new AWS services or exhibiting substantial shifts in their usage patterns. Proficiently detected and notified customer migrations with actionable insights to senior leadership.• Played a pivotal role in influencing a VP-level leadership decision to sunset an AWS service. Conducted a comprehensive data-driven analysis focusing on churn, adoption rate, fraudulent usage and customer migration journey, providing VPs with the insights needed to make informed choices about the future direction of the service ($360m 4w ARR)•Led the revamp of legacy ETL code logic, resulting in a 30-hour decrease in ETL processing time and a 43% overall improvement.•Implemented AWS Lambda Python scripts for automated ETL notifications and report generation, saving on-call monitoring time.

• Working as a BI engineer for EC2 core Product management team, an internal team that consumes data from multiple data warehouses (in TB’s/day) and computes usage/revenue, feature level data to provide internal AWS teams insights for over 100 AWS services. • Created first end to end data ingestion pipeline to ingest DynamoDB JSON data using Pyspark, AWS Glue to provide insights into features for AWS services. • Built and customized new metrics for the team to track new service/feature launch success such as source of the customer, organic AWS customers, ripple effect on other AWS services by these customers to quantify success and identify potential customers to target. Orchestrated the automation of these metrics into a user-friendly dashboard, with a template that garnered adoption across teams. •Identified and rectified two major COE’s caused from a legacy code which had been inflating metrics for a high-value service by ~20% ($54B, 4w ARR). Delivered a comprehensive impact analysis, and effectively communicated intricate code logic to leadership to mitigate their concerns.

•Parsed 20k JSON format resumes to improve the quality of data and applied machine learning methods to predict fellows’ duration in the program using Python to improve the admissions criteria and manage resources. •Prepared data science material and conducted 80+ interviews to select candidates into the program by assessing their technical and behavioral skills. Grew the Data Science track with 5-10 fellows joining monthly achieving a $500k+ projected annual revenue.

•Built and trained decision tree models, including XGBoost, Random Forest using H2O framework on time series data. •Forecasted sales for 1000 unique products across diverse regions and customers for Bridgestone.•Created a model that outperformed existing model with an accuracy of 82% and implemented the enhanced model for multiple clients. •Designed and presented visualizations to deliver comprehensive sales insights to VP-level leadership in a dashboard