May 2022 – Nov 2024 · 2 yrs 6 mos · Gurugram, Haryana, India · Hybrid

• Designed highly-scalable event-driven microservices for ship cost prediction using ML models to reduce the mean absolute error between the manifest and invoice amount by $0.30 for over 2 billion containers per month. This initiative provided more accurate data to customers and enabled them to make intelligent decisions related to profitability.
• Decoupled a pipeline for the ingestion and allocation of manifest ship costs, previously integrated with the invoice pipeline which helped in reducing the SLA from 18 hours to 6 hours.
• Developed a library to centralize filters and calculation of transportation legs. Integrated it across 4 services thereby reducing the data discrepancies by 90%.
• Identified and fixed a $30 million/month cost discrepancy sent to customers due to bug in the ETL jobs. Created a dashboard to monitor the key metrics and a testing strategy for rigorous validation as a part of change management.
• Discovered the cost drop of $50 million/month due to recoverable failures impacting the data coverage for multiple customers. Designed a mechanism to retry the failures happened in previous job runs which also reduced the manual on-call efforts by 5 days/month.
• Led the cost optimization initiative by analyzing monthly bills and identifying opportunities to improve the usage of resources, resulting in the monthly savings of $25K.
• Mentored 2 interns in developing a CI/CD pipeline and a ship cost tracker tool.

Jul 2021 – May 2022 · 10 mos · Gurugram, Haryana, India · Remote

• Developed a plugin to enhance the data ingestion workflow for the EDM application, enabling seamless downloads from Amazon S3 to a specified local workspace with advanced filtering options (e.g., creation date, prefix, and file type).
• Implemented a robust backup mechanism for the EDM console settings file, allowing automatic fallback to a stable state in case of corruption, thereby preventing application failure and improving reliability.
• Led automation of regression testing by enhancing the test harness to return and validate status codes against user expectations, eliminating two days of manual verification per test cycle.
• Improved overall code quality by leveraging SonarQube, resolving over 150 issues including bugs, security vulnerabilities, and code smells.
• Integrated GitLab API into the Build Radiator tool to visualize CI/CD pipelines across multiple projects, enabling leadership to track pipeline status and build freshness at a glance.
• Enhanced operational monitoring by integrating AWS ECS cluster data into the internal tracker tool using the AWS API, providing real-time visibility into cluster health and key metrics.
• Authored comprehensive documentation for three critical components of the Data Lake codebase, improving knowledge sharing and onboarding efficiency.