Projects

Used genetic algorithm to produce population of weights

Used inverse network loss as fitness function

Weights improved after several generations 

Hybrid model outperformed other classification algorithms

Used Tensorflow applied CNN to extract features from images with street views and use them to fine tune VGG16 model

Applied distance metric learning to retrieve images with satellite view but having same semantic information.

Model achieved 93% validation accuracy

Full stack web application deployed on Heroku

Used React.js for frontend, Django for backend

Used Postgres for relational database hosted on AWS. 

Features include add to cart, checkout and payment using PayPal API

Admin can add/edit users and add/edit items.

Check out the live application here!

Mined 12 million tweets from Twitter API related to QANON and 2020 US Elections

Performed Sentiment analysis using VADER 

Performed Topic Modeling using BERT

Conducted Location based Qanon user profiling

Mined 0.5 million tweets from Twitter API related to 2019 Canadian Elections

Extracted 100k unique users from the data set and calculated their attributes

Performed dimensionality reduction to yield 3 attributes

Applied K-means clustering with k=2 to isolate bot accounts

Identified 6% bot accounts and 94% Human Accounts

Mined 16 million tweets from Twitter API related to Brexit

Performed Sentiment analysis using VADER 

Performed Topic Modeling using LDA provided in GENSIM package

Positive correlation between GBP and Twitter sentiment

Extracted key discussion topics using unsupervised Machine Learning

Created a GUI in processing using Java 

Incorporated force feedback using Haply development kit 

Application allowed guided navigation with auditory feedback

Users feel greater resistance in densely populated regions on graph

Mined 1 million tweets from Twitter API related to Covid-19

Performed Sentiment analysis and Topic Modelling 

Public Sentiment was most negative in USA and neutral

Government was the main focus of discussion in all three countries