Jul 2024 – Dec 2025 · 1 yr 5 mos · Hybrid

• 1. Built a real-time Candidate Generator leveraging user’s short-term intent to power personalized product suggestions, enabling a widget-free, scalable feed-first homepage experience, this reduced Redis costs by 6,000 dollar/month which was used to store DS widgets data.
• 2. Built an in-session personalization pipeline for the FY feed, dynamically surfacing and ranking products based on real-time user interactions within the app session.
• 3. Developed a feature to integrate DS powered filters into backend system. This feature will help users to get to their desire products in efficient manner by providing accurate filters in segregate manner.

Jul 2022 – Jul 2024 · 2 yrs · Hybrid

• 1. Implemented a critical P0 feature introducing mid-feed filters, resulting in a 70 percent surge in filters adoption. Currently, in the process of scaling this experiment to encompass 100 percent of users.
• 2. Developed an in-memory caching layer over the Redis layer for the review and rating service, significantly reducing Redis IOPS by 75 percent and saving $2,420 per month in infrastructure costs.
• 3. Independently tackled a product task to display the number of orders sold on the PDP, resulting in a 1.5 percent increase in order-per-click ratio, indicating improved conversion rates.
• 4. Demonstrated proactiveness by voluntarily taking on-call responsibilities, diligently monitoring API and infrastructure metrics to ensure system stability.
• 5. Mitigated API abuse by implementing cross-site scripting prevention measures and enforcing size limitations on review comments to enhance security.
• 6. Developed a Python script to facilitate data publication from a CSV file to a Kafka topic, enabling seamless updates of product name templates on the Product Display Page (PDP) in HBase.
• 7. Implemented unit tests for key P0 services to improve maintenance, ensure robustness, and facilitate efficient debugging and future enhancements.
• 8. Enhanced monitoring capabilities by incorporating various metrics such as cache hit/miss analysis for in-memory caching, leading to improved system performance.
• 9. Designed and implemented a pipeline from scratch to extract rating data for a given product from Presto, ingesting it asynchronously into our MySQL database through Kafka using a cron scheduler.