Roopek Ravi

About Roopek Ravi

Data Scientist with 9+ years of experience architecting end-to-end machine learning and optimization systems for global leaders likeAWS and Ford. Proven track record of delivering over $50M+ in documented annual savings through innovative forecasting, causalinference, and resource optimization. Expert in building scalable ETL pipelines and deploying Generative AI solutions to automatecomplex operational workflows. Combines an advanced background in machine learning, AI, and statistical modeling to solve high-stakes challenges across cloud infrastructure, hardware reliability, and manufacturing.

Experience

1. Data Scientist

   Amazon Web Services (AWS)

   Apr 2021 — Present · Seattle, WA, US

   DynamoDB Capacity Team:• Architected an end-to-end machine learning-based capacity forecasting system(ARIMA, regression models and LSTM) for DynamoDB fleet management, analyzing historical usage patterns, seasonal trends, and peak demand events(Black Friday, Prime day, NYE, etc) to optimize server procurement, resulting in $10M+ annual cost savings in infrastructureexpenses.• Designed and implemented an intelligent multi-dimensional AZ balancing system using linear programming for IOPS, storage, and replica count optimization, increasing fleet-wide resource utilization by 15% while reducing capacity waste.Hardware Engineering Team:• Created a tool for the S3/Storage teams, enabling them to monitor and receive alerts on HDD drive failure rates based on various factors such as drive model, capacity, firmware, and region. The alerts have facilitated early detection of rising failure rates across 20 plus drive models, totaling more than 5 million plus drives in 2024.• Implemented a Bill of Materials (BOM) validation system that checks if HDD/SSD drive models are present in the appropriate servers during replacement of drives. The system helped reduce BOM mismatch during replacements by 99%.• Built a XGBOOST model that uses HDD/SSD SMART logs, device statistics and S3 diagnostic logs to determine falsely evicted/replaced drives. This helped the team reduce false drive replacement rate by 5%.• Designed and implemented global A/B testing and quasi-experimental methods to analyze environmental impacts on 2M+ servers, which led to the discovery of optimal temperature and humidity ranges for data centers, resulting inenhanced hardware reliability and an average decrease in failure rate by 6% across critical components (CPUs, motherboards, power distribution boards).

Education

• Arizona State University

  Master's degree, Industrial Engineering

  2014 — 2016

• Anna University Chennai

  Bachelor of Engineering (B.E.), Mechanical Engineering

  2010 — 2014