Welcome to My Portfolio

Autoencoders are a type of artificial neural network used for unsupervised learning. Their primary goal is to learn efficient representations of input data, typically for dimensionality reduction or noise removal. However, there is a lot more inherent in its architecture.

[20 minute read] • Artificial Intelligence (Intermediate) • Pre-requisites: Multi Layer Perceptrons and vanilla Autoencoders After this tutorial, you would be able to build your own multimodal generative model from scratch.

Developed two Conditional Generative Adversarial Networks, first one is CTGAN for tabular data and second one is Pix2Pix for image data to comic face. The project includes a web application for generating synthetic datasets using these models. The CTGAN model trains on the fly and generates synthetic data, while the Pix2Pix model is trained on NVIDIA P100 GPU for 9.5 hours on the Comic Face Dataset on Kaggle. The web app allows users to upload images and download generated datasets.

A web platform for campus students to find collaborators for their projects or a project to work on. Implemented major dynamic content in the frontend using React.js, reducing codebase and response time. Implemented a spam filtering algorithm using a heuristic criterion and reduced time taken by 67 ms per message.

A database design project for an Indian Army Camp, North Guwahati that supports daily local transactions in a single place while maintaining ACID properties. This system helps manage personnel, equipment, and operations data efficiently.

A fitness tracker and predictor application that uses both generative AI and classical ML algorithms for fitness-related predictions. Features include activity tracking, goal setting, and predictive analysis. Deployed and accessible online as of February 2025.

A real estate property search and management system implemented in C++. This application allows users to search, filter, and manage real estate listings with an efficient search algorithm.

A basic terminal-based text editor similar to Vim, implemented from scratch in C++ using pointers and linked list data structures. This project demonstrates low-level text manipulation and editor functionality.