Personal Projects

InnerCircle: A decentralized application on ethereum for creating pop-up events

• Developed as part of two-day hackathon organized by Consensys.
• Application developed using React, Solidity and Linnia protocol.
• Creating meaningful engagement similar to word of mouth by allowing anyone to create an event and invite trusted/loyal inner circle of members.
• Referral Mechanism to expand the circle.
• Try it yourself
• code is open source and available here

Booklyn: Decentralized e-book exchanges on top of ethereum

Completed @ NYU

Books2Rec: Book Recommender System For Goodreads

• As part of Big Data Science class by Prof. Anasse Bari.
• Hybrid Recommender System using Latent Factors from Reader-Book interaction and content based features for books metadata.
• Looking for recommendations for books to read, try our website

Secure Authentication system for mobile users

• A challenge-response based 2-Factor Authentication system using SMS channel
• Authentication backend in Python django
• Twilio APIs for interaction with mobile device via SMS channel
• Android application to interact with Authentication server

Forecasting average speed of buses between MTA Bus Stops

• Analyzed Time Series data of average speeds of buses between two stops among many such Stops
• Data cleaning and Preparation for Missing values in Time Series
• Model identification for each of the Time Series(AR, MA, ARMA ot ARIMA)
• Compared the Forecasting results with Linear Regression
• Spark was used for both Forecasting and Prediction using linear Regression

Classification of News Articles

• Achieved task of Classification of news articles(20-newsgroup dataset) into their respective groups
• Classified Using Naive bayes over tf-idf vectors of Bag of words
• Classification using cosine similarity of word2vec representation of documents

Completed @ Samsung Research India

Transmission Time Estimator for cloud applications in Smartphone

• Estimated time taken to upload/download media files on smartphone using Machine Learning.
• Wrote appllication for data collection in wild.
• Feature Selection to include most predictive features.
• Compared various Machine Learning models and selected Support Vector Regression for its superior performance.
• Presented work at IEEE WCNC 2017.

CPU Power modeling on Android Smartphones

• Improved CPU power model within Android phones for fine-grained power usage metric.
• Experimented to show inconsistencies with power reported versus actual power used.
• Collected CPU power data usage vs cpu frequencies.
• Formulated the problem as Quadratic programming and solved for new model parameters.

Data Traffic Manager

• Service to piggyback background transmissions with foreground date traffic in android.

Completed @ IIT Bombay

BattMan: Communication framework leveraging inaudible frequencies

• Studied GPS duty cycling techniques in crowdsensing
• Analyzed feasibility acoustic communication in Crowdsensing scenario
• Android application for sending and receiving messages using inaudible frequencies.

Simulator for Dynamic Server Allocation on virtual clusters using markov decision process

• Developed simulator to test Policies for virtual machine migration on a Data centre given migration costs and SLA costs.
• Optimized peformance of simulator using reduced state representation.

Automated Performance Testing framework for web servers

• Wrote services to add custom load patterns for performance testing of webservers.

Completed @ SICS, Sweden

Low power network protocol for sensor networks

• Data Transfer Protocol to transfer sensor data periodically from sensor nodes.
• Finite state machine implementation on Contiki OS for fault tolerance.