A collection of key projects in have done either in my free time or during Masters and Bachelors studies.

Data Science/Machine Learning

Includes Projects from courses taken at NC State University: Machine Learning for User Adaptive Systems, Algorithms for Data Guided Business Intelligence, Foundation of Data Science

  • [2016] Music Recommender System using Apache Spark and Python - Built a Music Recommender System using Collaborative Filtering approach on implicit feedback dataset with Apache Sparks’s Machine Learning Library (MLib) in Python.

  • [2016] Handwriting Recognition, Learning and Generation - Machine Learning based custom handwriting font generator. Applied character segmentation to handwritten documents, used MNIST dataset to train various Machine Learning models for character recognition, extracted key user specific writing features and studied application of RNNs for sequence generation. Report

  • [2016] Adwords Placement - Given that advertisement giants like Google, Yahoo only make revenue only if the viewer clicks the ads, it’s important to make the clever decision of which Ads should be shown (in limited number of ads slots available) depending on the query typed. Implemented Greedy, MSVV, and Balance algorithms in R for a very simplified version of this problem statement. MSSV beats other algorithms in terms of total revenue generated.

  • [2015] Anomaly Detection in Time Evolving Networks - Implemented Signature Similarity method for detecting Anomaly in time evolving graphs. Because of the way the similarity is calculated, anomalous graphs are identified by two consecutive anomalous time points in the output.
    Based on paper: Web Graph Similarity for Anomaly Detection

Web/Mobile Development

I really enjoy building web/mobile applications, whether big or small. Throughout my time in undergrad schoo, I lead the development of web portals for several technical events and clubs. My recent work experiences at Amazon and Morgan Stanley has further enchanced my skills in it.

  • [Ongoing] spamtest - Building a Django Python Web app for predicting if an email content is spam. Predicts the spamminess score based on a ML model pre-trained on huge dataset.

  • [2016] Context-aware Music Player (Android) - MusicRec is a prototype of a context-aware Android application which adapts at run-time bsaed on user’s activity. Song recommendations is based on whether user is driving, working-out, relaxing, etc.

  • [2013-14] Events Portal - Developed VNIT’s Annual Management festival (Consortium’12), and lead development of Annual Technical festival (AXIS’13) web portals. Handling thousands of participants’ data, optimizing webpages for faster load was bit of a challenge.

  • [2011-12] VLive - VNIT’s Social Networking Website - I feel most proud of building it from scratch using JavaScript, PHP and MySQL during the winter break of my sophomore year in undergrad school. 2 dedicated months, 5 team members and a bunch of awesome features like - forum discussion, facebook style chat, video call and notice board. Unfortunately it was hosted on internal college servers and hence inaccessible from outside.

    Mini Apps:

Talks