Apr 2013 – Oct 2019 · 6 yrs 6 mos · Greater New York City Area

• Started and led the Social Advertising Platform LOB (line of business) on the engineering front.
• Team: started the team with 5 engineers and grew to 35+ engineers with 4 engineering managers. I worked closely with the PM Team to understand the business requirements.
• Within the Ads LOB, I led building modules like Publishing, Automation, Analytics, Integrations, Optimization, Automation & External API.
• Infra & Platform: Architected proprietary plugins in ElasticSearch to support external lookups, joins and created highly efficient Compressed Cache. This improved performance by 3X. Delivered a very generic channel framework to make API calls to different social channels. This greatly reduced the error handling and maintaining the API connections.
• Publishing & Data Analytics: Architected "Data Analytics" module handling for millions of analytics records per minute. This involved exposing Webhooks to connect with the mobile/web analytics partners like GA, DCM, Kochava, etc. Build the model to publish Ads to channels like FB, Twitter, LinkedIn, (+ 5 others).
• Automation Engine, Budget reallocation: Delivered budget reallocation which became a very popular feature for reallocating budgets. Delivered rules engine to support millions of automated actions.
• Technologies Used: ElasticSearch (+Proprietary Plugins), Mongo, MySQL, Redis, Apache Mesos, Kafka, Spark

Apr 2008 – Apr 2013 · 5 yrs

• Terracotta is a distributed caching solution which could store petabytes of data in its cluster (peer to peer topology).
• Distributed Lock Manager (DLM): I was entrusted with the responsibility to deliver this module. DLM would allow super efficient locking mechanism across distributed machines. Many concepts like leasing of locks, compression, AA Trees, etc were used to make a very efficient system.
• BigMemory OffHeap: I was responsible for coding of the OffHeap Store on the server side. This would allow storing TBs of data in memory with GC pauses of less than 2 seconds.
• Availability: To achieve high availability, I helped in building the capability of Non Stop layer which would allow accessing data even when some nodes were down.
• Database Layer: Helped in designing and writing a database layer. This meant we could switch between our own proprietary database, Derby and Berkeley DB anytime.
• I joined as a Software Engineer and was Promoted as a “Team Lead” in 2010.

Jun 2006 – Mar 2008 · 1 yr 9 mos

• Worked in web conferencing team for BEEHIVE.