Attendbot

Smart Attendance Manager
This app is available on IOS app Store, with over 2000 downloads and a rating of 4.6 /5.0. This application aims to help students keep track of their attendance smartly, without much hassle reducing inputs. Using smart algorithms students don’t have to do much.

ChallengedTexting

This multiplatform application allows blind and deaf people to communicate using Morse code on a smartphone. Using vibrations on a smart phone, this application makes it possible for blind and deaf to type on a smartphone. The app is capable of doing both morse code to text and text to morse code conversion.

Minimum Spanning Tree

This has been project created as a Programming project submission for CSE -5311- 001 Design and analysis of algorithm. From the given project choices, I have chosen to implement and compare Minimum Spanning Trees Algorithm. This is a GUI python-based application that allows a user to perform create a Minimum Spanning Tree using either prims or Kruskal algorithm.

Lens IT , Object Recognition Application

This multiplatform (IOS and Android) application uses vision and trained neural networks models to recognise objects in daily life and provide a link to them in amazon or other popular online vendors.

Heavy Duty Vehicle Assistant

This multiplatform (IOS and Android) application was created with 3 of my classmates as our final college project. This applications main goal is to solve a problem which made transportation a hassle in India. Live tracking, Toll calculation , truck management , trip indicator and document requirement were some of the main features of the application. The application used several libraries like vision , NL by apple , realm db , GCP and many others.

Kaggle Notebook

Polynomial regression

This notebook implments polynomial regression on randomly created data using python and sklearn -library.

KNN on iris dataset

This notebook implmenets KNN algorthim from scratch using different distance caculation methods. Then KNN is applied on iris flower dataset to find the optimal hyperparamteres. Finally , the alogrithm is applied on testing data to calculate the accuracy of the model.

Multi-layer neural network from scratch Using Tensorflow 2.0

This notebook implmenets a multi-layer neural network using tensorflow 2.0 without using keras. This model can use only 3 types of activation functions, 'Relu' , 'sigmoid' and 'linear'.

NLP mulit text classification

This project applies state of the art machine learning technique, Neural Networks to create a multi class text classifier. It uses tensorflow.

Autonomous Vehicle System

This paper focuses on the performance of an autonomous vehicle system developed to run on devices with low computational power, lacking a Graphical Processing Unit (GPU). This system includes lane detection, speed calculation, steering angle prediction and object detection

Matrix Factorisation Algorithms For Highly Scalable Recommender Systems

In this paper, we experiment and compare different type of matrix factorization techniques for making recommendations for a recommender system using collaborative filtering. In particular, we will be using the SVD, QR decomposition, and Non Negative Factorization on a rating/ weight matrix.

Food Image Detector on Coco Dataset

This project applies object detection to COCO dataset using tensorflow and python. The model is a CNN network trained on batches of masked image datasets.