Savan Visalpara

Designed and implemented a CNN in PyTorch for the task of counting shapes in a given image. (PyTorch)

ThirdEye is an AI-powered device (i.e Glasses) that leverages Machine Learning techniques such as Image Captioning to help visually impaired people understand their surroundings better than before. Basically, this device narrates the surroundings for the users and also assist them in case of vague narration/caption. This project was awarded as the best project during the annual project fair at Alpha College of Engineering and Technology. [ TensorFlow, Keras, SciPy stack libraries ]

Trained multiple classifiers using Scikit-learn library to classify given text based on emotion.

Technology: Python, Pandas, Scikit-learn

1 - Implemented various GAN models such as vanilla GAN and DCGAN to generate face/mnist images. Purpose of this project was to learn more about various challenges involved in training GANs.
2 - Implemented Naive Bayes algorithm for semantic analysis on Twitter data. Also implemented it using different algorithms and measured accuracy of different algorithms. [Python, Scikit-learn, SciPy]
3 - Used Markov Models (first order/second order/third order/fourth order) to generate random text… Show more

1 - Implemented various GAN models such as vanilla GAN and DCGAN to generate face/mnist images. Purpose of this project was to learn more about various challenges involved in training GANs.
2 - Implemented Naive Bayes algorithm for semantic analysis on Twitter data. Also implemented it using different algorithms and measured accuracy of different algorithms. [Python, Scikit-learn, SciPy]
3 - Used Markov Models (first order/second order/third order/fourth order) to generate random text from given corpus.[Python]
4 - SEO ANALYZER: A command line tool in Python to check how SEO-friendly your website is. [Python]
5 - Multi Language Dictionary: Developed a GUI app using Python and Tkinter to get the meaning of the word in multiple languages. It uses concepts of web scraping. [Python, Tkinter, Beautiful Soup, lxml]
6 - Used item-based collaborative filtering to build a movie recommendation engine. [Python, Pandas, Jupyter Notebook]
7 - Trained a seq2seq model on Ubuntu dialogue dataset to develop a general purpose chatbot. [TensorFlow, Python] Show less

As part of the NLP class project, we were required to work on a COVID-19 related project. We decided to work on a BERT-based model that can extract multiple emotions from a text. We were provided with Reddit posts and the whole class helped to annotate them.
Finally, I trained a model on this dataset to identify emotions.

Mastering a word often requires some effort, one of the standard procedure for this is to write down a word and revise periodically. Vocabulary Master comes with a GUI app, Chrome extension, and a web app.
In this project, I've tried to make this process digital. Vocabulary Master lets you add, search and practice words anytime you want. To make the process of mastering a word more convenient and useful, along with this GUI app and web app, I have also created a Chrome extension which will… Show more

Mastering a word often requires some effort, one of the standard procedure for this is to write down a word and revise periodically. Vocabulary Master comes with a GUI app, Chrome extension, and a web app.
In this project, I've tried to make this process digital. Vocabulary Master lets you add, search and practice words anytime you want. To make the process of mastering a word more convenient and useful, along with this GUI app and web app, I have also created a Chrome extension which will show you a word from your word database each time you open a new tab.

Technologies : Java, JavaFX, JSP, Sqlite, JavaScript, Jquery, JSON Show less

Trained a Machine Learning model to detect malicious URL using scikit-learn. Used TfidfVectorizer and trained multiple models with various settings of parameters. Also developed a web app using Flask for demo purpose.

Developed a web application as a part of government's hackathon to make students record available online.

Technologies : PHP, JavaScript, HTML, CSS