Aravind Kumar

900 High School Way Mountain View, CA 94041 · (978) 798-2202 · karavindkumar1993@gmail.com

Hi Am Aravind Programmer,Hacker,Love for Functional programming
I am a Backend Engineer (Machine Learning) at VMware experienced in building high performance,mission critical and time sensitive applications.

  • Experience

    Member of Technical Staff

    VMware
    • Improved write speed of the system to the database by 80% by :
      • Removing bottlenecks caused by locks among different systems competing to write to a single database by parallelizing the writes from different sub-systems to its own individual databases.

      • Improved the scalability of the system by integrating the pacemaker tool thereby adding the ability to add and/remove additional machines on the fly without loss of data and performance

      • Reducing the number of write operations to storing results in memory and returning the stored result rather than re-computing the result for every request.

      • Removed the bottleneck between the database thread and system thread by use of Lock-Less Ring Buffers as a form of communication between threads.

      • Offloading serialization of data using Protocol Buffers from Main Thread to each of subsystems designated threads.

    • Implemented Batch-Reads and On-Demand reads on the High-Availability System thus :
      • Transforming the system from a read-only/write-only system to a simultaneous read-write system

      • Thereby reducing the downtime of the system in the event of a failure from 3-4hours to 0 hours (> 95% improvement).

        • Implemented data mirroring in real time (Hot-Standby) with the backup system which enables to read and/or recover rows on the fly from the primary system without any downtime resulting in 50% reduction in downtime in the event of a failure

    Senior Software Developer

    Sandvine
    • Improved write speed of the system to the database by 80% by :
      • Removing bottlenecks caused by locks among different systems competing to write to a single database by parallelizing the writes from different sub-systems to its own individual databases.

      • Improved the scalability of the system by integrating the pacemaker tool thereby adding the ability to add and/remove additional machines on the fly without loss of data and performance

      • Reducing the number of write operations to storing results in memory and returning the stored result rather than re-computing the result for every request.

      • Removed the bottleneck between the database thread and system thread by use of Lock-Less Ring Buffers as a form of communication between threads.

      • Offloading serialization of data using Protocol Buffers from Main Thread to each of subsystems designated threads.

    • Implemented Batch-Reads and On-Demand reads on the High-Availability System thus :
      • Transforming the system from a read-only/write-only system to a simultaneous read-write system

      • Thereby reducing the downtime of the system in the event of a failure from 3-4hours to 0 hours (> 95% improvement).

        • Implemented data mirroring in real time (Hot-Standby) with the backup system which enables to read and/or recover rows on the fly from the primary system without any downtime resulting in 50% reduction in downtime in the event of a failure
    Fall 2015 - Spring 2017

    Software Developer

    Sandvine
  • Devised the Emergency Call Support feature for Rx and Gx Diameter Applications of “Policy and Charging Rule Function (PCRF)” node of a wireless network which involved:
    • Creation of separate message queues and processing units to prioritize the Emergency Call Traffic.
    • Adding Support for Custom Policy Expressions/Hooks
    • Adding a Command Line Interface (CLI) framework.
  • Fall 2014 - Spring 2015

    Side Projects

  • Opensouce Contribution:
    • Contribute to Red-Hat Pacemaker Project:
      • Added Stonith-Max-Attempts feature into Pacemaker thereby enabling user to configure the number of attempts to be made for stonith.
      • Added support for on the fly debugging for a resource and its resouce-agents
    • Sandvine Hackathon : Integrated Distributed Compiler (icecream) into the Sandvine Build System.Resulting in 84% improvement for each compilation.
  • .

    Skills

    Programming Languages & Tools
    Skills
    • C++
    • C
    • Java
    • Python
    • Android App Development
    • Javascript
    • HTML
    • CSS

    Education

    Northeastern University

    Master of Computer Science
    Computer Science - Algorithms,Machine Learning,Operating System,Functional Programming

    GPA: 3.5

    Fall 2017 - Fall 2019

    Amrita University

    Bachelor of Computer Science
    Computer Science - Algorithms,Data Structures,Compiles,Data Mining,Cryptography,Functional Programming
    Fall 2010 - Spring 2014