Flows by MoEngage - Campaigns and Notifications• Designed an event-driven finite state machine based architecture for Flows to track user behaviour and visualize end-to-end customer journeys.• Developed Apache Samza based real-time streaming jobs. This system processes 15K user events and 5K user attributes per second. • Developed the backend for Flows for campaign creation and management.• Reduced the memory usage of Flows streaming job by 65% and increased throughput to 10X.Data Platform• Core member of the team responsible for upgrading Kafka from v0.8.2 to v1.1.0.• Upgraded all streaming pipelines from Samza v0.11.0 to v0.14.1 and migrated them to new Hadoop YARN cluster, while maintaining consistency and reliability, without any downtime.• Set up and maintained Hadoop YARN clusters for all streaming pipelines.• Set up Druid + Superset based monitoring systems for streaming pipelines.• This system handles more than 3 billion events per day combined across all Kafka topics.Awarded "Rising Star" for outstanding performance for the term 2017-18.Tech Stack: Kafka, Apache Samza, Hadoop, MongoDB, RocksDB, AWS

Developed ETL pipelines to aggregate different types of data from various sources to serve SCM Inbound in-house applications including analytics dashboards, pricing management, advertisements and buy-assist systems. Developed keyword ranking algorithm to identify top trending products based on scraped data from competitors and fashion based media.Worked with the pricing team in the Data Science department for dynamic pricing, adaptive discounting, MRP prediction of in-house brands, scripting and developing ETL pipelines.

Worked with the Seller platform team to develop a machine learning based guidance tool for the product Sell for Me, aka eBay Valet.

Worked on applying Dijkstra's algorithm to very large graphs using OpenMP.