Supporting Analysis of Dimensionality Reduction Results with
Contrastive Learning
• Implemented the paper with the same title as the title of
this project. The main aim of the project was to learn data
science techniques to visualize data. Apart from the
visualizations in the paper, two more visualizations were
added to get more insights and to explore relationships and
patterns in data.
The frontend was developed using HTML, Bootstrap, CSS,
JavaScript, and D3js was used for data manipulation and to
visualize data. Backend had a novel algorithm “ccPCA” which
is a variant of PCA and is used to contrast clusters. The
backend was implemented in Python and the whole project was
hosted on Flask server.
A Flask-based Web App designed to exemplify the power of
PaasS platform by deploying this scalable application on
Google App Engine. It uses GAE, Microsoft Speaker
Recognition API, Google Speech-to-Text API and Cloud
Firestore.
I worked on a Machine Learning project to detect cells with
cancerous symptoms. Implemented Deep Neural Network
algorithms VGG-16, VGG-19, ResNet and CNN. I made 16
variants of these algorithms with varying architecture and
the algorithms had an accuracy to detect cancer between 52%
to 97%.
Project demonstrating a web application to manage blogs.
This project demonstrates how to integrate a React Web
Application and integrate it with Google Firebase to make an
application that ensures realtime updates and employs a
serverless architecture.
A complete deep-dive into VueJS from scratch. Contains
content and code from very basic concepts of VueJS to
complex VueJS features like Vuex and Vue Router.
A complete deep-dive into Node.js from scratch. Contains
content and code from very basic concepts of Node.js to
complex Node.js features like Express framework, MVC
Architecture, Routing. Also contains content on making
RESTful and GraphQL APIs.
Implemented an autoscaling project that allowed edge
computing on Raspberry PI (IoT device) which allowed object
detection. Used darknet to perform object detection. The
autoscaling was done using our code and not using Amazon’s
Elastic Load Balancer.
The traffic was directed to Amazon EC2 instances, which were
directly scaled up and scaled down based on the number of
requests waiting in the Amazon SQS queue. All the object
detection results were then stored in Amazon S3 buckets.
Analysis of Geospatial Data
Performed mining of geospatial data in Phase 1 using Hadoop
and HDFS. In Phase 2 I used Apache Spark and an extension of
it that is used to manipulate and mine the geo-spatial data
“GeoSpark” to find the “Hot zones of NYC Taxi” which shows
at which locations are these taxis available and “Hot cells”
which had geo locations with most number of taxis at a given
point in time using Getis-Ord Statistics. The code was
written solely in Scala.
L”Earn”
Implemented an interactive website that contains all the
programming tutorials pertaining to my university’s
curriculum. This was a personal project.
Front-end and back-end were developed using Java Server
Pages (JSP), JAVA and database was managed using ORACLE 10g
DBMS.