Software Engineer Ii
CurrentEstimation platform • Designed and developed the Browse Node estimation platform which determines the accuracy of correctly assigned products to the browse leaf node destinations in the Amazon catalog. • Designed and led the development of automated calibration workflow which gets the training data, post-processes and trains/calibrates the models on a regular basis for up-to-date accuracy. This helped save manual efforts and reduced the calibration time from a week to a day. • Helped onboard and mentor an intern to develop the calibration workflow and produce a quicksight dashbaord for displaying the weekly metrics and provide insights. • This system helped track and improve the precision of top products from approx 90% to 95% across all countries.On-demand Reclassification system • Designed and developed a single stop solution ODC - a self-service capability to fix misclassification defects using machine learning (ML) classifiers. • Designed and developed several REST APIs for clients to be able to ingest products, integrated with BN classifier models which run at 90%+ precision to get optimal classification and integrated with the correction service to perform the correction. • Around 250M products were classified by this system out of 325M input products in the year 2024. Languages: Java, Python, ScalaAWS Services: Lambda, Step function, Batch, EMR, S3, Glue, DynamoDB, Athena, Quicksight