Projects and Work Experience

AWS ElasticBeanstalk

Currently working at Amazon Web Services on ElasticBeanstalk team as a Software Development Engineer. Learn about ElasticBeanstalk here.

Vein-to-Vein

As a Master's student I was working on C4G Vein-to-Vein an open source software solution for tracking the lifecycle of blood from the point of donation to transfusion. My advisor for this project was Dr. Santosh Vempala. This software product is aimed at replacing the manual procedure adopted for managing blood products in several developing countries of Africa. C4G Vein-to-Vein started as a collaboration between the Computing for Good (C4G) program at Georgia Tech, the CDC, Safe Blood for Africa and the participating countries of Africa. Initially developed at Georgia Tech, this project is now being maintained by Jembi under the name Blood Safety Information System (BSIS).

Safe management of blood is of primary importance to prevent Transfusion Transmitted Infections. Once blood is collected at a site it is packed in a collection bag and quarantined. Meanwhile a sample of blood is sent for testing. If the test results indicate the blood is safe for transfusion then it is made available for use. When a clinic sends a request for blood of a specific type and quantity to the storage center, only the blood bags marked as safe should be issued. Unsafe blood units are disposed off. The management of this information is the responsibility of Vein-to-Vein. Vein-to-Vein is an affordable product tailor made for the needs of developing countries. At present we are collaborating with labs in Zambia and Cameroon.
Online version is available for Demo at http://bit.ly/C4G-V2V.
(Username: test, Password: test)

Follow this project on Github:


Internship at Facebook

I interned at Facebook in the summer of 2012 as a Software Engineer in the Datacenter Power Modeling/Optimization team. I was reporting to Qiang Wu during the internship. A site like Facebook requires a lot of servers to run. Very often servers need to be replaced and new servers need to be added to ensure high uptime requirements. These decisions need to be taken with an eye towards energy efficiency and power safety limits imposed by the hardware. I was part of the team responsible for creating visualization and emulation tools to make this decision making easier.


Work at Oracle

Between the time I completed my undergrad at IIT Roorkee and started as a grad student at Georgia Tech, I worked as a Software Developer for an year at Oracle Server Technologies, India in the Directory Services team. The team is part of the Identity Management group at Oracle and creates several LDAP compliant products. I was reporting to Satishkumar Venkatasamy. Specifically I developed on a virtualization module for Oracle Unified Directory. Patent Filed:"Directory Server Processing Requests Based On Hierarchical Models While Using Backend Servers Operating Based On Relational Models" Filed on 22 May 2012.


Undergraduate Thesis

For my undergraduate thesis at IIT Roorkee I worked with Akhil Langer under the guidance of Dr. Ankush Mittal. The title of our thesis was "Providing Natural Language Interface to Healthcare Services Over SMS". We aimed at researching novel ways to provide access to health related information over the SMS medium. SMS though constrained to 160 characters are fairly popular and inexpensive means of communication. At the end of the project we developed three applications:

  1. Find a Doctor: Given an unstructured SMS query containing the address, health problem description and optionally the name of a doctor, find the closest doctor for the described health problem.
  2. Drug QA: This system was designed to fetch information about a medicine. The user can phrase question about medicine in natural language like "What are the side effects of Aspirin?", "When should I take Ambien?" and the system would return the relevant answer from a database. We trained our system on a database of questions captured from health forums. We handled misspellings by using spelling correction libraries.
  3. FAQ Search: Given a question in SMS Lingo this system would search a corpus of questions for the most similar question and return the answer for the same. We experimented with a corpus of questions downloaded from Yahoo Answers. For this system we used a combination of a combinatorial search scheme and a syntactic tree matcher to get an improvement in the prediction accuracy over the previous systems.
Our project fetched us the Best B.Tech. Project Award in Computer Science from IIT Roorkee for the year 2010. We also won the Innovative Student Projects Award, 2010 from Indian National Academy of Engineering for our project.