Modern Business - Start Bootstrap Template

Hosting a static website on AWS S3 bucket without making it public.
Saved all the front-end files on an S3 bucket, and then created an AWS Cloudfront network and gave it access to the bucket's contents.
Purchased a URL from AWS Route53, got a certificate for it from AWS Certificate Manager
Attatched that certificate and the URL to the cloudfront network.
Built a CI/CD pipeline using Github Actions, as it will make it easier to make any changes in the website.

A full-stack website to review campgrounds created by others all over the world and add your own campgrounds for others to review.
Front-end created using HTML, EJS, CSS, and Bootstrap 5.
The database used is MongoDB Atlas to store campgrounds, users, reviews and sessions.
Node.js and Express.js are used for server side application.
For user authentication Passport.js is used.
For the authentication of the data entered by the user Joi Schemas are used.
All images are uploaded to Cloudinary via Multer.
All the maps are taken from MapBox API.

A React e-commerce web application to sell planter pots.
Hosted via AWS Amplify Gen 2.
AWS Cognito is used for authentication.
This project effectively displays the use of state, passing it from parent to child as props, and using context api to provide state to multiple children all at once
Used react-router-dom to create routes for different components.

A Node.js web application just like Google Drive where you upload files and generate password protected links to share and download them.
Implemented text file compression using the Huffman algorithm, reducing file sizes by up to fifty percent.
Leveraged MongoDB Atlas for efficient data storage and it is hosted on AWS EC2 with PM2.

A tensorflow dual encoder model that is trained to detect outdated comments in Java code bases
Trained a fasttext model to generate embeddings for all the code-comment change samples in our dataset.
First we train the two encoders to consider the cosine similarity between old_comment (comment in previous commit) and old_method (method in previous commit) as 1
Case 1: train the two encoders to consider the cosine similarity between old_comment and new_method (method in new commit) as 0
Case 2: train the two encoders to consider the cosine similarity between old_comment and new_method as the cosine similarity between old_comment and new_comment (comment in new commit)
Results: Rank-error=0.3564, F1-Score=0.026, Recall=0.019, Precision=0.042
Imp. Libraries used: Tensorflow, Numpy, Pandas, FastText, Keras

Uniview gives you a summary of google maps reviews for a particular university that you are interested in.
Utilizes AWS Lambda to fetch and filter 1,500 reviews from the google maps profile of 100 different universities.
Performs sentiment analysis on each review using AWS Comprehend and uses AWS Personalize to recommend 5 similar universities for each university.
Implements TF-IDF NLP algorithm to extract 10 most common words from positive and negative reviews.
The entire project was built on Terraform. Other services used: EC2, Amplify, S3, Cloudwatch, DynamoDB.

CUP is a Seq2Seq model and HebCUP is a heuristic-based approach, both are used to update outdated comments.
We divided 1,063K code-comment pairs between CUP and HebCUP labels based on the NCSM scores.
We aim to update each code-comment pair based on its label and compare our performance with using either CUP or HebCUP alone on the entire dataset.
Imp. Libraries used: tensorflow, numpy, keras, pytorch, cudatoolkit, cudnn, scikit-learn, scipy

Projects