San Jose, California, United States
*Everything I write on LinkedIn are my own views and do not reflect that of my current employer or any past employer I was associated with* I graduated from the University of Southern California. My area of interest lies in building highly-scalable infrastructure. I worked as an Infrastructure Engineer at Hike Messenger, India's youngest unicorn startup. All this while, my responsibilities had been to understand the product requirement and deliver it on time, working with the team to write clean, reusable pieces of code, and build scalable architecture for microservices, and mentoring new hires.
• Collaborated with data science team to develop a personalization engine, producing required data on the messaging queue (Kafka) and integrating API’s in user flow which increased Click Through Rate (CTR) by 30% • Architected and implemented a microservice using Java to deliver actionable post to user’s timeline and storing their actions in HBase. Exposed this service through thrift endpoint.
• Coded a gatekeeper system which tracked notifications sent to the user on hourly, daily, and weekly basis in HBase and controlled outgoing notifications based on the data from HBase and smart rules, this reduced spam by 13% • Implemented Redis & Kafka based queuing system which produced data to Kafka when Redis was unavailable while also maintaining the packet order, later switching back to Redis when Redis became available, thereby improving the reliability of infrastructure by 5% • Kafka Upgrade: Led initiative to upgrade Kafka from 0.10.x to Kafka 1.1, completed upgrade activity on production system with zero downtime • Programmed a search system and indexed a 100 million users record in Elasticsearch that improved end-user experience for the search of others. • Enhanced and ensured consistent user experience across networks by transcoding video in multiple resolutions for different devices and serving high-quality video based on multiple parameters.
• Setup pipeline and dashboard for viewing analysis based on user action logs in real time with ElasticSearch, Logstash, Kibana (ELK) stack which improved the visibility of infrastructure. • Reduced memory required for storage of user segments by a factor of 64 by storing segments in bitset instead of MongoDB documents.