Oct 2018 – Apr 2023 · 4 yrs 6 mos · Menlo Park, California, United States · On-site

• Data Infrastructure – Graph Analytics
• Led the engineering execution for DIGraph/Giraph (Meta’s distributed in-memory graph engine), overseeing an 8-engineer team across multiple pods to improve efficiency, security, scheduling and service excellence.
• Curated the technical vision and re-architected the Graph Compute engine for heterogeneous server types, reducing power consumption by ~30% (~1.5 MW).
• Architected and implemented a privacy-aware integration and shadow testing framework, automating release cycles of multiple engines and saving 40–60 developer hours per week.
• Developed scheduling protocols for DIGraph, improving platform efficiency by 20% (~1 MW).
• Designed and exposed Graph abstraction API on DIGraph, increasing platform adoption from 50 to 200+ teams.
• Implemented core components of Meta’s cross-org Infra Demand & Capacity Control framework, enabling accurate cost estimation, admission control, and capacity planning across a dozen orgs.
• Implemented end to end authentication and authorization of the open-source Apache Giraph platform.
• My (limited) open source contributions here include:
• https://github.com/apache/giraph/commit/2c63aa2321a570f7dfcc659f565d18afb63a2e3c
• https://github.com/apache/giraph/commit/1acb587f807a8e7b0e50bfd19b85a246db9e2abd
• https://github.com/apache/giraph/commit/d3bf4a2cf5347f7cfd9d217b216c906cb7801217

Feb 2017 – Sep 2018 · 1 yr 7 mos · Menlo Park, California, United States · On-site

• Ads Delivery and Product
• Designed features such instream-ads reserved buying, contextual ads, Audience Network Integration for the “Reach & Frequency” (optimized for predictive reach and controlled frequency) product. This helped bring in additional $XB+ revenue per year for Meta.
• Implemented “waterfall logging” across the entire Ads web tier, which helped identify poorly performing code blocks. Optimized these which resulted in 40% improvement in request latencies perceived by advertisers.
• Optimized delivery for multi-media rich ads, working across entire ads delivery and indexing systems.

May 2016 – Aug 2016 · 3 mos · Menlo Park, California · On-site

• Ads Delivery and Product
• Developed an analytical model (classifier) in C++ to classify under delivering ads. The work involved framing and implementation of rules by correctly fetching, analyzing and storing the required data.
• Exposed new API’s in Hack (PHP) to call the above classifier via Thrift.
• This classifier was used to display and track under-deliverying campaigns, both to internal and external users, and possible reasons for the poor performance, which helped advertisers take necessary action to improve ad delivery.

Jun 2014 – Jul 2015 · 1 yr 1 mo · Bengaluru, Karnataka, India

• Designed and developed an accounting system to track orders, returns and refunds on Flipkart.com. This system created invoices for every transaction and ensured there were no inconsistencies in accounting. The data from this system was used to generate financial reports at the end of each month.
• Developed the new PDF templates for Invoice and payment receipts of Flipkart.com.
• Technologies: Ruby, Sinatra and MySQL

Jul 2013 – Jun 2014 · 11 mos · Bengaluru, Karnataka, India

• MySQL Server Development Team
• Designed and implemented the user-level password expiration policy in MySQL using C++.
• https://github.com/mysql/mysql-server/commit/48f3e6654d5039a18fbaaa51c844622f124cd3fe
• Developed a feature to bring MySQL server into offline mode, which ensures only super users can connect to the MySQL server and gracefully disconnects other clients. https://github.com/mysql/mysql-server/commit/efe34b41b3d3618a09799cfa558d82fee0fead76
• Built a prototype for replacing the regex library of MySQL, with a modern library.

Jun 2013 – Jun 2013 · 0 mo · Bengaluru, Karnataka, India

• Implemented a tool (in Python) to sample S3 datasets from large ones, based on certain filters.
• Added support to the Qubole SDK to download results from S3 location.
• Extended the Qubole SDK to add support for the scheduler API.

Feb 2013 – Apr 2013 · 2 mos · Bengaluru, Karnataka, India

• Developed a module in Python to collect parameters that are to be monitored, to help build the Hyper-V driver for Healthnmon (a monitoring tool for OpenStack Virtual Machines).
• Technologies used: Python

Jun 2012 – Aug 2012 · 2 mos · Bangalore

• Developed a few automation tools which will help database administrators perform their day to day activities in a much easier and faster way. The tools developed include Database Snapshot Backup Status Viewer, Crontab Editor/Viewer and SQL-ID Comparator.
• Technologies: PHP, AJAX, HTML.

Jun 2011 – Aug 2011 · 2 mos · Bangalore

• Developed a multi-lingual application; for the ManageMind video surveillance software; which helps the user in performing all types of administrative operations right from installation, backup and restoration of database, viewing and exporting logs etc in a user friendly manner, thus increasing the overall efficiency of the ManageMind software.
• Technologies: Core Java, Java Swings, MySQL, XML.