CV

Education

• B.E. in Computer Science and Engineering, Sri Venkateswara College of Engineering, 2013-17
• M.S. in Computer Science, University of Massachusetts Amherst, 2017-19

Work experience

• Summer 2018: Graduate Research Intern, IBM Research
  • Created an entity relationship extractor using piecewise-CNN modified with positional and contextual features, on unstructured news data. Used Watson NLU, Watson Knowledge Studio for pre annotation task, PyTorch for model and D3.js for visualisations
• Fall 2018: Graduate Student Researcher, Microsoft Research Maluuba
  • Performed visual reasoning on a synthetic dataset (FigureQA) with 76% accuracy. Implemented a Relation Network with modules for FiLM, self and group attention to answer relational questions on graphs using PyTorch.
• Summer 2017: Software Engineer Trainee, Global Analytics India Pvt Ltd.
  • Deployed Android app to visualize data from company reports using custom classes to analyze and plot 7 types of graphs and charts. Improved efficiency of calculations by implementing new interfaces.
• Fall 2015: Data analytics intern, Ramjay&Co
  • Performed analysis of tax and income data of clients using Hadoop and MySQL.
• Summer 2015: Research Assistant, College of Engineering Guindy
  • Implemented Naive Bayes classifier to perform Sentiment Analysis on a stream of twitter data using Hadoop, reaching 74% accuracy (then on par with state of the art). Overcame the reducer slow start problem in Hadoop by modifying the reducer to start without waiting for the mapper to finish.

Skills

• Programming
  • Python
  • Java
  • C++
  • MATLAB
  • C
  • MySQL
  • Cassandra
• Frameworks
  • PyTorch
  • TensorFlow
  • Caffe
  • Android Studio
  • Hadoop

Projects

• Diagonolytics- Predictive analysis for preventive medicine
  • Predicted insurance claims from medicaid data(CMS.gov) using linear regression
  • Predicted future diseases with 79% accuracy using 10-nearest neighbor classifier
  • Analyzed similarity of diseases, and predicted severity by k-means clustering of patients data, and used Tableau to display
  • Won ‘Best Healthcare and AI hack’ from IBM Research, ‘Best Open Source Hack’ from Two Sigma
• Visual place recognition from Google StreetView images
  • Developed and compared multiple models to predict coordinates of a StreetView image
  • Detected cities first using SVM-CNN on extracted features(GIST, color histogram, raw pixels) to select from models trained on different cities to predict the coordinates of image
  • Achieved a classification accuracy of 98% and haversine distance of 39km for locations in the United States
• ‘Novella’- a story weaver bot (Hack UMass)
  • Developed a chat bot which generates a coherent story conversationally with both text and voice
  • Structured the bot with Amazon Lex, integrated responses using the AWS Lambda function to connect with a Hidden Markov Model using trigrams which was trained on 7 books, to synthesize responses
  • Interfaced with an Android Application which provided both text and voice responses
• Interactive Image Description to assist the visually challenged
  • Led a team to develop system to answer questions based on images using Convolutional Neural Nets and Long Short Term Memory
  • Used TensorFlow, NumPy, pickle and 15 other libraries in Python and a web application in Sinatra
  • Achieved an exact answer match of 61% and relevant answer match of 86%
• Smart Bin: SPI Hackathon
  • Finalist in “SPI Hack for Social Good” for developing a Smart Dustbin to segregate recyclable from non-recyclable waste
  • A camera and motion sensor connected to Raspberry Pi detect placement of waste and send image to a server
  • Responsible for building a Neural Network on Octave to process this and return the classification which then rotates a motor on the platform accordingly so it falls in the right bin
• Intbot: A conversational AI agent
  • Integrated a chatbot with the official app for Interrupt, a National Level Symposium, to answer participants’ queries about events; enhancing user experience and allowing redistribution of workforce
  • Synthesized responses using Artificial Intelligence Markup Language(AIML) and modification of Program AB integrated with Android
• Prediction of term deposit based on historical bank data: CTS Hackathon
  • Built decision tree classifier to predict if a customer would take up term deposit based on attributes like credit score and other data
  • Utilized the Rattle library in R to tune the decision tree and visualize it for enhanced feature engineering
  • Won 2nd runner up with an accuracy of 92%

Teaching

COMPSCI 197C - Special topics in C programming

Honors and Awards

• Best Healthcare and AI Hack, Best Open Source Hack *IBM at SheHacks Boston
  • Developed dashboard to assist physicians in preventive healthcare, which provided general analysis of patients data and additional predictions about future diseases based on present conditions of patients along with side effects of prescribed drugs, predictions about possible claims in the future, etc
• Second place *KURUKSHETRA’16, College of Engineering Guindy
  • Second place at Heptathlon, which tests 7 core areas in Computer Science - Algorithms, Data Structures, Networks, Operating Systems, Databases, Web technology, Theory of computation
• Second runner up *CTS Ltd Hackathon
  • Prediction of term deposit based on historical customer data related to the theme of ‘Banking and Insurance’ using Decision tree in R. Built UI with Sinatra.

Organizations

• Vice Chair, ACM Student Chapter, SVCE, 2016-17
• Head of Academics, ACM Student Chapter, SVCE, 2015-16
• Head of Design, ACE, SVCE, 2015-16
• Head of Design, Literary and Dramatics, SVCE, 2015-16