The project,done during my internship at InfoChimps comprises a set of widgets written in the Ext JS 5.0.0 Sencha framework for Infochimps' new Internet Of Things initiative, in particular to enhance the Map component of the IOT application. The widgets provide a number of nifty, custom overlays for the map using several open API's like Google Maps, Google Earth, Open Weather Map etc. You can see the repository here.
Checking NBA teams popularity on Twitter using Accumulo. I used a ready made VM with Accumulo already installed and wrote a shell script for automating login to the accumulo shell for various operations.The sentiment analysis is extremely rudimentary, simply checking for 'win' or 'lose' in tweets. It was more to learn Accumulo than anything else. Check out the repo here.
Used Feed-forward Neural Network and error back propagation with SciPy to train a machine learning program to accurately recognize and predict handwritten digits. The project used the well known MNIST dataset and you can view the code here.
Using Map-Reduce to recommend friends to Social Network Users.This project implements a simple “People You Might Know” social network friendship recommendation algorithm. The key idea is that if two people have a lot of mutual friends, then the system should recommend that they connect with each other. View the source code here.
Used Hadoop MapReduce (2.6.0) to efficiently compute the 10 most and least volatile stocks on the NASDAQ over the last three years on datasets containing key metrics for upto 29000 stocks. Also repeated the experiment using Pig and Hive(abstraction layers built on top of Hadoop). You can see the GitHuB repository here.
Comparing various Classification and Regression Techniques in Machine Learning.We examine a number of techniques like Linear/Quadratic Discriminant Analysis, Logistic Regression, Ridge Regression and assess their effectiveness in terms of RMSE (root mean square error) programmatically using Python/SciPy. View the source code here.
This was my first project as an offshore consultant to Citi Equities while working for Larsen & Toubro Infotech Ltd.Part of a team that supported over 90 Unix servers (Linux and Solaris) as well as around 30 database servers (Sybase ASE 15, Microsoft SQL Server 2005) globally which hosted Citi Equities investment banking applications in the QA/UAT environment
My day-to-day responsibilities included performing various application checkouts, bash shell scripting work for automating various support tasks, monitoring and maintaining of servers and databases using support tools like ITRS, deploying and sanity testing new builds, troubleshooting and fixing application issues (with or without developer intervention depending on severity), applying patches and being first point of contact for coordinating solutions to a variety of tech issues on a 24X5 rotational basis.
I did this project while consulting for the Citi Trading Analytics team.The trading analytics team uses kdb+ a high speed database that achieves superior performance by storing vast amounts of real-time and historical data in memory. This data can be retrieved quickly and 'crunched' by an investment banking application in order to help make a risk based decision about a trading position. The database is queried and programmed using a language called Q.
I was responsible for support and testing of the trading analytics application in the QA environment including deployments, smoke testing, functional QA, L1 and L2 troubleshooting and investigation of any issues in QA, patching daily checkouts etc as well as liaising with development and business users to resolve issues involving any service/process.
Implemented a 5-way client server system using C on Unix platform in which 4 individual department machines acted as clients/peers with the ability to register with a fifth server as well as transfer(either upload or download) files among themselves and/or the main server. All communication was carried out using TCP sockets.