Close

OpenRepositoryScore - GitHub Sentiment and Toxicity Analysis

---

A frontend only React Server Side Rendered (SSR) app to analyze toxicity and sentiment in open source repositories on GitHub

Application Details:

Tools Used

• Tensorflow.js with pre-trained LSTM based Toxicity and Sentiment Analysis models
• Next.js with server side SSR - hosted on now.sh
• Auth0 for simple SSO authentication

Details

Close

Karana Health ID

---

An app developed during Stanford's Health++ Hackathon. Created as an offline first Electron Health Record (EHR) system.

Application Details:

Application: Microservice Architecture


Services:

• React frontend SSr app to render QR codes with vital health data.
• Hasura GraphQL Service to store speech information
• Auth0 for simple SSO authentication

Details on Devpost

Close

Awaaz - Machine Learning for Speech Impediment Detection

---

An app developed during Calhacks 6 to detect speech impediments such as stutter and aphasia. Inspired by Google's Parrotron Blog Post. This app was inducted into NSF's I-Corps Bootcamp Program. Currently under development. Reach out if you are interested in applied machine learning for healthcare!

Code:

Application: Microservice Architecture
Services:

• Python Flask api + deployed classification ML model
• React frontend app
• Hasura GraphQL Service to store speech information
• Auth0 for simple SSO authentication

Close

Kiva.org - Crowdfunding Analysis

---

Extensive exploration of Kiva.org's crowdfunding data on Kaggle

Code

Close

Mumbai Amber Alert

---

A simple chatbot based missing person alert on Facebook Messenger. The project is currently a part of Facebook's Global FbStart program under the Bootstrap and Social Change track.

The project won the Facebook Bots for Good Hackathon. Complete details here.

Close

Aspect Based Sentiment Analysis

---

Minerva is a Ruby on Rails applcation at the core of the sentiment analysis system.

Code:

Presentation

The entire system comprises of:

1. Aspect Extraction Module - Minerva

   A Ruby on Rails application that performs infrequent feature mining and opinion mining on a single movie review. It then resolves the noun phrases extracted (as features) using Carnegi Mellon University's NELL Knowledgebase.
   After resolving phrases to higher level 'concepts', we make multiple parallel async calls to the sentiment analysis module to get sentiment scores and labels. All this information is then stored inside a Redis Store in a logical tree format.
   This information is then visualized using D3.js (for the logical aspect tree) and plotly.js (for the sentiment scores).

2. Facebook Messenger Chatbot - SilverScreen

   A bot built using the Sinatra framework. Makes calls to Minerva and receives and results and forwards them to the user. The user interacts via Facebook Messenger's Buttons and the movies are fetched from the TMdb API. Once a movie is selected, the bot scrapes NYTimes for uncached movie reviews.

3. Sentiment Analysis Module - Eleos

   The sentiment analysis module uses two machine learning approaches to analyse the text input fed by the context resolution system. One is a simple NaiveBayes classifier for review classification. The other is the implementation of doc2vec model which generates relative vectors of word embeddings. After the infrequent features extraction is performed on the NYT movie review text, each opinion-sentence pair is sent to this sentiment analysis module. The module deals with one system identifying the classification of the input text. The other system gives a score to the text which denotes how strong the sentiments behind the input text are. The datasets used for training are the IMDB movie reviews for review classification.

Close

Railway Time Series Analysis

---

• The data was collected through the planitmetro web site, which hosts several datasets provided by the Washington Metropolitan Area Transit Authority data. This data has been cleaned up -stations which did not have any information about time were removed from the database. The dataset consists of over 28000 rows, each representing a group of 5 time peaks for a single station, identifying the number of average rides each day at 5 distinct times of the day (AM_PEAK, PM_PEAK, MIDDAY, LAATE NIGHT PEAK, EVENING). The data is accumulated for a period of approximately 6 years from Sept 2010 to Jan 2016.
• To understand the 'distance' or correlation between stations, we increase the distance between two stations based on the overall variation between the %change in Passenger Density. For example, if a Station A has 2 data values (Passenger Density) - 4000 & 5000, it means an increase of 25%. Similarly, Station B having 2 data values - 400 & 500, also has an increase of 25%. Thus, both these stations have the same variation and thus could belong to the same cluster. Initially, we pre-process the data and compute the distance matrix.

• I used Agglomerative Clustering on the distance matrix to find 10 clusters (default), and plot the Passenger Density with time.
• I compare the no of clusters with the cluster cohesion measure (Silhouette score & Calinski-Harabaz Index). The results are quite interesting, the cohesion measure itself falls with an increase in the number of clusters.
• When the number of clusters is 2, they show the maximum cohesion and when we look at the heatmaps, it becomes easy to understand why. The heatmaps have multiple lines of a darker shade (higher distance measures) cutting across a background of a lower shade (low distance measures).

Application: Jupyter Notebook running Python 3 kernel.
Tools Used:

• Pandas
• Matplotlib with Seaborn
• Sckit Learn modules:
  1. Agglomerative Clustering
  2. BIRCH
  3. Silhouette Coefficient
  4. Calinski-Harabaz Index
  5. Numpy

Close

VoicePass - Passwords as a Service

---

Close

Uber Projects

---

• I built a couple of web applications for the Uber Onboarding centre in Mumbai during Summer '16.
• The Queueing system I built to call out tokens provided a more convenient way to handle sign ups and registrations. It was really challenging to scale since it was a real time application built on PusherJS.
• It also had Admin functionalities and queried Papertrail logs for daily metrics via Sidekiq Workers.

• Application Types:
  1. A Real Time, Multithreaded web application
  2. Apps Script
  3. Google Sheets API
• Technologies & Tools used: Sinatra, PusherJS, Redis (for Sidekiq Scheduling), Memcached (saviour in terms of scaling the real time system), ToastrJS (notifications), PaperTrail (logs and data analytics)

Close

ELK Stack with Docker and Rails

---

Code:

• This was an implementation of the well known ELK (Elasticsearch - Logstash - Kibana) stack on Docker with Rails.
• I wrote Dockerfiles for each part of the ELK stack, the Rails app (a simple scaffolded app) and the Nginx Proxy. I tied it all up with a single docker-compose.yml file to bring up the containers.
• Since Rails logging by default is verbose, to store it in Elasticsearch, it needed to be configured to output logs in GELF (Graylog Extended Format).
• Once configured, the Logstash source was configured to be Rails logger output, and the destination Elasticsearch. Kibana ran on the root endpoint of the server and the Rails app ran on /railsapp served via Nginx.
• Thus, the user can access the Rails app on the /railsapp/posts and Kibana on /. The Ruby on Rails app sends logs to Logstash via 12201/udp port (GELF input endpoint) and Kibana and Logstash talk to Elasticsearch via port 9200 (JSON REST API port). Nginx serves as a reverse-proxy server and web server and routes requests to Kibana or the Rails app based on the endpoint that you visit. The application was deployed on a AWS Lightsail server with 2GB memory, 1 CPU and 40GB SSD.

Close

Heroku Flock Integration

---

Code:

Built during the Flockathon held at Directi Offices, Mumbai. My team and I built a simple chatbot (similar to Slack's slash command) for the Flock team communication app.

The slash command allows you to authenticate accounts, setup and deploy Heroku pipelines as well as view the last 25 releases.

Close

Popular Movies Android App

---

Source Code:

App Demo on Appetize

• Built as part of the Udacity Android Nanodegree
• Specifically, the purpose of this project was to learn how to build clean and compelling user interfaces (UIs), fetch data from network services, and optimize the experience for various mobile devices. The app communicates with theMovideDB API and provides a responsive and delightful user experience

• Application Type Android Application (>API 22)
• Features
  1. Uses the Picasso Image Caching Library
  2. Performance enhancements such as using ViewHolder patterns, Parcelable etc

Close

Study of SQL and NoSql Datbases in a distributed environment

---

HuddleInc is a website which promotes Community created and curated events. The Role of the project was to compare the scalability of SQL and NoSQL Databases in a Distributed Environment.

• Databases Used: Postgresql & MongoDB
• Developed as the Semester 6 Distributed Databases(CPC603), Software Project Management(CPE6012) and Software Engineering(CPC602) Project
• Application Type: Ruby On Rails Web Application
• Further Details:

Close

Implementation of the Apache OpenNLP Library using JRuby

---

Just some quick hacking with Jruby and the OpenNLP Libary.

Java has one of the most extensive libraries. While I love Ruby's gems, JRuby provides access to the JVM coupled with Ruby's amazing syntax. Win Win

• Application Type: Command Line Program. Input & Output : Text Files
• Implemented as part of a Process Oriented Guided Inquiry Learning (POGIL) project under System Programming and Compiler Construction (CPC601)
• Source Code:

Close

EduHub

---

The winning entry for the HackForIndia Hackathon organized by Digital Vibes.

Problem Statement: Use the IBM Watson API to solve a societal problem among the given domains.

We picked the education domain and created an application which converts typed notes into audiobooks. This way, the audio files can be downloaded and listened to during commute and distributed to people who do not have access to quality education as well.

• Application Type Ruby On Rails Web Application using the Watson API
• Source Code:

Close

Unity 5 Games

---

I learnt Unity 5 and many aspects of Game design by building 3 games.

• Application Type Unity 5 Browser Based and Windows 7 Games
• Details
  1. Roll A Ball 3D: Learnt Basic Setup, Moving the Player and Camera, Creating Collectable Objects & Scoring.
  2. Space Shooter: Learnt Camera, Lighting, Game Mechanics (Collision, Speed etc), Creating Game Objects with Textures (ROcket Flame, Shots etc), Setting up Boundaries, Dynamic Effects (moving rocks, explosions), Hazards, Scoring etc
  3. Survival Shooter: Learnt Player Health bar, Status, 3D movement, Spawning Enemies, Game Mechanics etc.


• Source Code:

Close