AgileBot

Project for CSC510 - Software Engineering

Mentor: Prof. Chris Parnin
Team Members: Omkar Acharya, Yogesh Lele, Amit Watve, Jalandhar Singh

• AgileBot is a SlackBot that interfaces with Project planning tool Rally. When called for its services, AgileBot looks up for unassigned bugs and user stories, estimates their time frame and assigns the user stories to the 'best' engineer after taking into consideration various heuristics.
• AgileBot's mission is to automate these repetitive tasks to a degree that will help engineers stay in flow by reducing distractions, eliminating context switches and thus increasing productivity.
• @AgileBot can be added to the team's Slack Room.
• For backlog grooming, the Scrum master can mention @AgileBot with text GroomBacklog to automatically assign points to the backlog stories.
• For sprint planning, a Team Leader can invoke AgileBot by mentioning it to PlanSprint.
• Engineers working as a part of the sprint can submit their everyday statuses by mentioning AgileBot with the text GiveMyStatus.
• Scrum Master/Team Leader can view the entire team's status by calling GetTeamStatus.
• AgileBot fits into the Space Reactor category.

Want to take a look at the project? Check it out on GitHub

Close

Complete DevOps Pipeline

Project for CSC519 - DevOps

Mentor: Prof. Chris Parnin
Team Members: Omkar Acharya, Rishi Jain, Shriyansh Yadav, Urmil Parikh

• Milestone 1: Configuration and Build Management
  We provisioned a Jenkins server on an AWS EC2 instance. We created a build job on Jenkins server using Jenkins Job Builder. This job will perform the build commands and if successful, the post-build scripts will be run for each application. As a part of the post-build scripts, we provisioned 2 AWS EC2 instances (one for each application).


• Milestone 2: Test and Analysis
  We implemented 3 sub tasks viz. Commit Fuzzer, Test Prioritization Analyzer, and Automated Test Generation. For commit fuzzer, we used mutation-based fuzzing We made 100 different fuzzed commits and ran test suites on the code base. After those 100 runs, we prioritized the test suites based on the time required for the test each run and the number of times the test failed. These two sub tasks were focusing on iTrust application. For checkbox.io, we developed an automated test generator which generates test cases for every API route in checkbox.io.


• Milestone 3: Deployment and Infrastructure
  We deployed iTrust using Rolling Updates strategy whereas checkbox.io using Canary Release strategy. We created a dockerized version of checkbox.io and deployed it on Kubernetes cluster. We also implemented a feature flag using Redis which will enable/disable a particular feature of checkbox.io when toggled using a simple web application.


• Milestone 4: Special
  We implemented 3 sub tasks. The first one is Docker Monkey. We deployed 3 EC2 servers (one as a proxy server and 2 production servers). When the CPU usage on any of the production server exceeds the threshold, the proxy server stops forwarding requests to that server. We also included ESLint as a part of post commit hook. As the third task, we implemented a load handler which makes the production server Read-Only when it is flooded with POST requests.

Want to take a look at the project? Check it out on GitHub
Want to explore more about this? Check out our presentation

Close

Yelp Restaurant Photo Classification

Capstone project for CSC591 - Algorithms for Data Guided Business Intelligence class

Mentor: Prof. Nagiza Samatova
Team Members: Omkar Acharya, Amit Watve, Akshay Arlikatti

• Implemented a three-stage transfer learning pipeline with Convolutional Neural Nets and Support Vector Machine
• Extracted features from with 234,842 training images and 1,190,225 testing images using BAIR Reference Caffemodel
• Trained a Support Vector Machine (SVM) One vs All classifier using these features for multilabel learning
• Ranked 134th out of 355 in the Kaggle competition with an F1 score of 0.764 on testing data (0.797 on validation)

Check out the presentation
Want to take a look at the project? Check it out on GitHub

Close

Settlement Mapping

Capstone project for CSC522 - Automated Learning and Data Analysis class

Mentors: Dr. Ranga Raju Vatsavai, Krishna Gadiraju (TA)
Team Members: Omkar Acharya, Amit Watve, Sridtt Bhalachandra, Felix Kim, Sean Mahaffey, Jonathan Raynor

• Developed an unsupervised learning system of exploring the settlement activities for planning and disaster recovery
• Implemented Gaussian Mixture Models (using Expectation Maximization algorithm) and 1-Holt Rule Based classifier
• Achieved ~85% accuracy on the validation set containing the different patches from a satellite image of 176MB
• Evaluated its performance with Weka’s RIPPER (Rule Based Classifier) and Scikit-Learn’s in built GMM model
• Designed and presented a poster along with a paper of the project workflow and its implementation details

Want to explore more about this? Check out the poster and the paper for this project
Want to take a look at the project? Check it out on GitHub

Close

Sentiment Analysis

Mini project for CSC591 - Algorithms for Data Guided Business Intelligence class

Mentor: Dr. Nagiza Samatova
Team Members: Omkar Acharya

• Built a sentiment analysis mini-project using the database containing the real-time tweets and IMDB reviews
• Approach 1: Logistic Regression with important words as features (Accuracy: Twitter 60.3%, IMDB 84.8%)
• Approach 2: Doc2Vec with Artificial Neural Nets and document vectors (Accuracy: Twitter 63.56%, IMDB 83.49%)
• Used Twitter’s Streaming API to fetch and analyze the tweets at runtime and stored them using Apache Kafka buffer

Want to take a look at the mini projects from CSC591? Visit my GitHub repository

Close

Personal Health Manager

Project for CSC540 - Database Management Concepts and Systems class

Mentor: Prof. Kemafor Ogan
Team Members: Omkar Acharya, Amit Watve, Akshay Arlikatti, Nathan Faulkner

• Developed a system to assist people to manage the health information, recommendations and emergency alerts
• Used Java Database Connectivity (JDBC) to interfacing with the Oracle SQL 11g database containing the tables for patients, health history, their health supporters, and diseases
• Implemented multiple triggers in Oracle to notify the patient and his/her health supporters about the unusual event

Want to take a look at the project? Check it out on GitHub

Close

VidScribe - Video to Text

Senior year group project at iKnowlation Research Labs

Mentors: Dr. Parag Kulkarni, Prof. Sheetal Sonawane
Team Members: Omkar Acharya, Parag Ahivale, Nehal Belgamwar, Gurnur Wadhwani

• Developed a system to describe the video content in English using Convolutional and Recurrent Neural Networks
• Finetuned two of the Nvidia’s CNN caffemodels (Google-Net and VGG-Net) on ImageNet dataset of 1.2M images
• Trained Long Short-Term Memory (LSTM) models on MSCOCO dataset of 120K images with Chainer framework
• Evaluated the performance of the model and finetuned the hyperparameters to improve its accuracy from 63% to 77% based on the existing conventional systems
• Won the 3rd prize for paper presentation (out of 35) at Birla Institute of Technology, Goa

Want to explore more about this? Check out the report and the paper for this project
Want to take a look at the project? Check it out on GitHub

Close

Leave Manager

Junior year project for Employability Skills Development Lab

Mentor: Prof. Pravin Patil
Team Members: Omkar Acharya, Parag Ahivale, Nehal Belgamwar, Anmol Agarwal

• A leave management portal for the faculty members in the undergraduate institution

Want to take a look at the project? Check it out on GitHub

Close

ACM ICPC - Regionals

ACM International Collegiate Programming Contest - Mid Atlantic Regionals 2016
Coach: Prof. David Sturgill
Team Members: Omkar Acharya, Austen Adler, Jacob Dremann

Close

Machine Learning

Coursera's course on Machine Learning by Stanford
Instructor: Prof. Andrew Ng

Close

Mining Massive Datasets

Coursera's course on Data Mining by Stanford
Instructors: Jure Leskovec, Anand Rajaraman, Jeffrey Ullman

Close

Web Application Architectures

Coursera's course on Web Applications by University of New Mexico
Instructor: Prof. Gregory L. Heileman

Close

Computer Science 101

Stanford's course on Introduction to Computer Science
Instructor: Prof. Nick Parlante

Close

Introduction to Linux

Stanford's course on Introduction to Computer Science
(Self-paced course)

Close