Health FundIt | Cryptocurrency Miner for Health Funding
Health Fundit is a desktop application built using
electron.js. The app works as a background process when started. After regular intervals,
number of hashes mined and some other stats are sent on the web via the REST API. As
per requirements of the client I configured the mining pool and formatted the API GET
and POST requests along with some statistical graphs in the application.
  electron.js
  
  JavaScript
  
  REST APIs
  
Share It | Hate Speech Detection
Share It is a CMS with automated hate speech and offensive language detection. My contributions
in this project included building a classification model for classifying blogs and other
textual data. The model had an accuracy of 84% on test data. I also had to integrate
the trained model with the php-backend of the CMS.
  Python
  
   Machine Learning
  
  Sentiment Analysis
  
  NLP
  
Big 5 Personality Traits Identification
Built a Web Application using
Flask for identifying the "Big 5 Personality Traits" based on a set of questions
and Twitter Stream of a user. User had to answer all questions and provide his/her twitter
handle. Twitter handle is used to extract tweets of a user using
Tweepy . Tweets are then preprocessed and fed into a Machine Learning pipeline which
outputs a probablity score for each of the 5 traits. I created a custom Machine Learning
pipeline involving NLP techniques like TFIDF vectorization and Stemming of tweets.
  Web Application
  
  Flask
  
  Twitter API
  
  NLP
  
Staff Trace
Staff Trace is a Desktop Application built using
electron.js for managing the Staff members of a staffing company in California. Staff
members are located using zip codes they operate in and are shown by a marker on Google
maps. Users can search for staffing services by entering thier zipcodes and other details.
  Javascript
  
  Google Maps API
  
  Management System
  
AI based automated Spell Checker for Urdu Language
Built a spell checker for Urdu, based on classic NLP techniques and algorithms like Ngram Frequencies, Soundex, Least Common
Subsequence etc. After a week of fine tuning the NLTK classifiers and the whole workflow,
I achived around 90% accuracy on test corpus.
  Python
  
  NLTK
  
  Urdu
  
  Algoithms
  
Data Science & Machine Leaning Notebooks
I have worked on various freelance projects involving traditional Data Scince and Machine Learning workflows using different
Datasets. Most of the work includes Data Cleaning, Exploratory Analysis, Predictive Modeling
and Story Telling.
  Pandas
  
  Numpy
  
  Matplotlib
  
  Scikit Learn
  
  NLTK