• Assessment of Twitter as a Review Platform

  Performed statistical and sentiment analysis on tweets to formulate custom indices. These ranking indices helped to determine whether Twitter can be used a sustainable and viable platform for products and services review and rating. Analyzed the various factors using Tableau's visualization.
  Tools & Technologies: Python NLTK, Cassandra, Apache Spark, Tableau, Twitter API.

• Automobile Recommendation System

  Developed a recommender system for automobiles based on indirect User profile, Car specifications and semantic analysis of reviews and news available for the car using Natural Language Processing implemented in Python.
  Tools & Technologies: Python NLTK, SQL.

• House Price Prediction and Crime co-relation analysis

  Analyzed the various factors that influence the House prices using regression and other models. Established correlations between the Spatial-Temporal influence of Crime rates.
  Tools & Technologies: Python Scikit Learn, QGIS.

• Wolfpool: An easy way to plan your rides

  Managed and Designed the Amazon’s Elastic Beanstalk PaaS for the application implemented using Node.js and MongoDB. It is a responsive web application that will facilitate planning and sharing of car rentals/Uber/Lyft for people based on Spatial and Temporal rule-based system. This will be useful in places where UberPool and LyftLine are not available
  Tools & Technologies: AWS, Node.js, MongoDB, HTML, jQuery, CSS.

• Microdata API for Servo, browser engine from Mozilla Research:

  Contributed to the Servo open source project by designing algorithm to extract machine-readable metadata in web content using Rust, programming language. This metadata is used to create vCard/vEvent file with the help of servoshell embedding.
  Technologies used: Rust, HTML.

• Capstone: Retrieving, Processing, and Visualizing Data:

  Developed a web-crawler using Python and “Page-rank” algorithm to create search engine index, stored in SQL. Designed a process to extract data from a large email dataset. Performed mapping and cleaning operation on the data to analyze the trend of various dimension like sender’s domain, word frequency, email count. Visualized the inbound/outbound links in form of connected graphs, histograms, and line graphs using D3.js library.
  Technologies used: Python, D3.js.

• Machine Learning by Stanford University:

  Implemented Image recognition, Clustering, Anomaly detection and Recommender systems using regression, classification, Neural-Networks, Support Vector Machines and other Machine Learning techniques. These algorithms were implemented in Octave.
  Technologies used: Octave.

• Optimization of Traveling Salesman Problem using Neural Networks:

  Implemented a hybrid approach that comprises an Adaptive Neuro-Fuzzy Inference System and Simulated Annealing algorithm to determines the path based on distance and other factors like profit, the probability of successful deal etc. Published journal articles in IJCSE about the theory and implementation of this project.
  Technologies used: C#, HTML, JavaScript, Google Maps.

• Data Warehouse and Mining Tool with ETL:

  Performed extraction of data from different file formats, data cleaning, and loading the consolidated data into MS Access database. Implemented various data mining algorithms like Naïve Bayes, K-Nearest Neighbors, K-Means Clustering and Hierarchical Clustering on the consolidated data using Java.
  Technologies used: Java, MySQL.

• Attendance tracking solution:

  Built a technology-enabled solution, a ‘Real-Time Attendance Tracker’ for one of the largest not-for-profit organizations in India known as ‘Teach For India’ in less 24 hours. This paperless attendance tracker was implemented using IVRS and SMS at its core along with a lightweight mobile app. This application was for the JPMorgan Code For Good Challenge, Mumbai 2013.
  Technologies used: HTML, CSS, PHP, MySQL, JavaScript, Exotel SMS API.

• AngryArmy:

  A 2-D Java applet based game, the goal of the game is to destroy opponents castle by use of cannons. The projectile motion vector of the warhead is impacted by wind direction and speed, the accuracy of the impact is used to calculate the damage on opponent's castle.
  Technologies used: HTML, CSS, Java Applet.