Deepak Singh

Led team responsible for managing lifecycle of machine learning models for Amazon's BuyBox algorithm, driving enhancements in prediction accuracy and operational efficiency.
Model Training Pipeline: Designed and maintained Big Data workflow that generates model features from historical data, enabling effective training of FMA models.
Model Tuning Workflow: Developed robust system for tuning existing model parameters on regular refresh schedule, improving model performance and alignment with business goals.
Model Deployment/AB Testing Framework: Spearheaded deployment of trained models to Tier-1 system, ensuring smooth integration and operation within critical Amazon services.
Model Evaluation: Led evaluation of A/B experiments involving BuyBox, providing actionable insights that informed strategic decisions and model iterations.

Led the Buying Option Analytics Data team, responsible for managing the data lifecycle for the BuyBox algorithm on Amazon's retail pages, directly impacting the efficiency and accuracy of product listings.
Managed data collection, processing, and governance processes, ensuring compliance with industry standards and internal policies.
Developed and implemented data management strategies, enhancing data quality and accessibility, which contributed to the success of Amazon's retail operations.
Collaborated with cross-functional teams to ensure data integrity and streamlined access control, supporting the broader goals of Amazon's retail platform.

Led the migration of the FMA training data collection process from the Detail page to a new stack, in alignment with a company-wide service migration initiative.
Ensured data quality and continuity throughout the migration process, maintaining the integrity and reliability of the data training pipeline.

Worked as full stack developer and lead Migration of Amazon Music detail page to new service stack.

Worked on prototyping a web-based tool for commodity trading using AngularJS and a microservice backend with an event bus system. This also involved integrating data from legacy software

Worked on an online reconciliation application for banking transactions. Created the backend using Spring Boot and REST/HATEOAS principles, and used Spring Security for authentication

Worked as a tech lead for a project management tool, primarily involved in integrating it with other projects in Zycus using ActiveMQ as the message broker

Started as an intern for 3 months, then worked on a project management tool that used JavaScript/JSP for the frontend and Spring/Hibernate with Oracle for the backend.