Introduction


I am Varun Manjunath a 2nd year Computer Science Master's Student at The University of Colorado,Boulder. My areas of interest include Machine Learning as well as Full Stack Web Development. In addition to that I enjoy playing Table Tennis and Badminton.

Linkedin
Github
Varun

Skills


Kubernetes
Docker
Ansible
Machine Learning
Deep Learning
React
Flask
Python
Java

Projects

Brewery React App

A brewery react app that implements search functionality. The user can search the breweries based on the location, address, state,city and number

Project
project screenshot
Viz-Wiz Visual Question Answering System

Designed a Viz-Wiz Visual Question Answering system that takes an image and question as input and outputs the answer to the question. The question is related to the content present in the image. The model utilizes a combination of Resnet50 for image feature extraction and a bidirectional LSTM for question encoding. Achieved an accuracy of 44.2 percent on the validation set

Project
project screenshot
Music Store Simulation

Utilized Object Oriended Design Patterns in order to build a simulation of a music store. The user will be able to choose how many days the simulation will be ran for. As the simulation is running, announcements of all actions that occurs during each day will be printed to the terminal.

Project
project screenshot

Experience

  1. project screenshot

    QI Path -- Front End Developer

    Working on migrating the front end from PHP to Angular for the QI PATH system that is used to find, prioritize and address vulnerabilities in any project.

  2. project screenshot

    Arista Networks -- Software Engineering intern

    • Learned about the Code Base which revolves around the concept of creating an Agent. There is a centralized database called Sysdb which the Agents connect to in order to exchange information. Agents are reactive in nature, which means that when an input state changes an action is taken.
    • Unblocked the VTI, VNI policer CLI in Arfa so that Microsoft can configure it One. This was a requirement by Microsoft although the feature is not supported on the switch. Wrote unit tests to validate functionality
    • Worked on MLAG (Multi-Chassis Link Aggregation) switch where I had to Refactor code to sync only Flood List entries across MLAG switches. This was an optimization from the previous design that utilizes VcsSde to sync Flood List entries. VcsSde has many unnecessary fields that are not a part of the MLAG sync. Wrote Unit tests and Integration tests to validate functionality.

  3. project screenshot

    VMware -- Applications Administrator

    Responsible for deploying and maintaining applications on platforms like PCF(Pivotal Cloud Foundry) and TKG(Tanzu Kubernetes Grid). Also maintain and troubleshoot Harbor, an application that is a repository for Docker images

  4. project screenshot

    VMware:Contract -- Data Science Intern

    • Developed a Context-Based Recommender System that provides persona-based content to VMware employees on the Intranet based on the Employee’s function, role, location, and interests.
    • Around 15GB of data was pulled from the VMware data center (Santa Clara Datacenter) and stored locally.
    • For every user, a set of 10 articles were recommended by taking a combination of Collaborative and Content-Based filtering in Recommender systems.
    • The Recommender system achieved a personalization score of 0.744 which is pretty high. The personalization score measures how customized the recommendations are for a set of users.

  5. project screenshot

    MindStix Software Labs -- Data Science Intern

    • Built an Adaptive Learning & Collaboration Platform to provide requirement-based training to beauty advisors of the Estee Lauder cosmetic company. The platform varies the difficulty level of questions based on the answer that the user provides.
    • There were a total of 30 questions and these questions were equally categorized into Easy, Medium, and Hard levels. Each level of difficulty has points associated with it with Hard being the most points followed by Medium and then Easy. On average, the users of this industry got 66% of the total questions correct.
    • Built a Chatbot application to deliver information about cosmetic products to Estee Lauder cosmetic company. The Chatbot utilized an open-source machine learning framework called Rasa NLU to train on intents and entities.
    • Around 200 question-answer pairs were collected from company employees and trained on the Rasa NLU framework.