MADHUR KAPOOR
Student - Data *Enthusiast - Ever Learning

Know Me!
Resume

Hello There! Welcome to my humble abode on the internet.

I am a Computer Science graduate student at the University of California, San Diego, specializing in Data Science, Distributed Systems, and Machine Learning.

I am enthusiastic about:

  • Java and Python Development
  • Large scale data analytics
  • Healthcare analytics, Adtech
  • Hadoop ecosystem and other distributed
    processing systems
  • Data Science and Visualization
  • Scikit-Learn
  • Fullstack (MEAN) Development
  • Budding Javascript Libraries
  • Trekking & Traveling

My Skills

Java

Python

Hadoop Ecosystem

MySQL

Javascript

Node JS

C++

C

PHP

Powershell

Couch DB

Linux

DATA SCIENCE, MACHINE LEARNING, ACADEMIC RESEARCH

My Work So Far

    Yahoo! Inc.

    (Intern)
  • Built an ad-inventory forecasting model for guaranteed ads based on stochastic streaming algorithms (sketches) which had a 3-fold improvement (accuracy, disk-space, infrastructural simplicity) over the existent sampling based system.
  • Collaborated directly with the tech lead and architect on this project which culminated with a paper submission in Yahoo! Tech Pulse 2016.

    San Diego Supercomputer Center

    (Graduate Student Researcher)
  • Analyzed census-level data across 5 years to aid public health ocers in data-driven policy-design decisions. My work was funded by the Robert Wood Johnson Foundation.
  • Developed a data pipeline for extracting the most meaningful factors aff ecting health outcomes like diabetes and life expectancy and their evaluation by comparing them from a theoretical stand-point. Involved similarity analysis, feature engineering/selection schemes and predictive analytics.

    Qulinary

    (Part-time Freelancing Software Engineer)
  • Working on an Ionic & Cordova app for real-time driver tracking and route optimization using AngularJS, Leaflet.js
  • Implemented an on-demand route optimization service to suggest an optimal order of traversal to drivers and aid them in deliveries.

    Citrix R&D

    (Software Test Engineer I)
  • Developed automation for the Personal Virtual Disk and Personalization product line from scratch up to 42% to reduce testing time of each build by minimizing manual work.
  • Designed and implemented UPM Troubleshooter (released as a hotfix), for customers and admins to debug their UPM setup which significantly reduced the round-trip-time for addressing customer issues.

    Artoo

    (Software Engineering Intern)
  • Engineered an Admin dashboard for the client - Ujjivan Financial Services, to track the activity of their field agents in providing loans to Bottom of Pyramid entrepreneurs.
  • Incorporated a lazy module loader into the application to significantly improve the boot-time.

    Indian Statistical Institute

    (Research Intern)
  • Devised a model to predict spurious citations in research work submitted to a conference proceeding. The estimations were made using text processing methods like TF-IDF.

Projects

Distributed File System

Implemented a distributed le system from scratch powered by a self-made RMI Library to support remote calls. Supported operations include CRUD les, naming, auto-replication, locking to ensure consistency. Know more

Predicted the top pick-up zones at any given hour, using 6 months Uber pickup data from NYC. Know more

Uber Analysis
Amazon Review Ratings

Predicted the helpfulness of a review (regression) and rating that a user would give to an item (using latent factor models)

Handwriting based collaboration framework-multiple remote parties collaborate using pen and paper in real-time. Click to know more

Hand Written Remote Content Collaboration
Eye D

Handwriting based collaboration framework-multiple remote parties collaborate using pen and paper in real-time. Click to know more

A QR Code based system to provide educational solutions for students who can't afford provate tutoring Click to know more

QR Code
In-Organization Search Engine

Mini search engine to query over textual data of an organization and return the related documents, using a single node Hadoop cluster

WORK EXPERIENCE