Kartik Aneja

Worked on a project under Mozilla Builders Open Lab program. Developed a web-based application for relief coordination on Twitter using machine learning models to identify and match resource request and offers. The website lists donation/request tweets location-wise, based on search.

Implemented Naive Bayes Classifier for classification of tweets(Donation/Non-Donation, Donor/Requestor & Resource Type classification), with an accuracy of 80%, after parsing them using standard NLP techniques.

Website: https://help-for-all.herokuapp.com/

Teaching Assistant for course ESE 344 Software Techniques for Engineers (Data Structures and Algorithms in C++) under professor Dr. Murali Subbarao

Worked on TI-MSP 430 micro-controller. Built a Hall Sensor-based Digital Odometer for Bicycles using Texas Instruments microcontroller MSP430. Designed and fabricated the PCB for the product and performed routing for the same. Programmed the controller in Embedded C.

Developed a Blog application using Django which lets users share posts on the forum. All users can view each post on the blog by other users. The web application allows users to create, update, and delete their own posts. The application supports sign-up and sign-in feature and provides authentication using Google.

Built a print server on my local network using Raspberry Pi and Common Unix Printing System to provide wireless connectivity to the printer over the network.

Built a speech recognition algorithm that understands and converts simple speech commands into text using 1-D Convolutional Neural Network. The training data is taken from TensorFlow Speech Commands Datasets, which includes 65,000 one-second long utterances of 30 short words, by thousands of different people. Achieved an accuracy of 82.95% on test data.

Built a Simple and Multiple Linear Regression Model to fit and predict house prices on Boston Housing Dataset and NYC Airbnb Dataset. Assessed the model using Residual plots with parameters like co-variance, linearity, and normality. Performed Feature reduction using the Hypothesis Testing (p-test) and co-relation among features.

Trained the AlexNet Convolutional Neural Network for the purpose of image classification by fine-tuning the last layers. Achieved a prediction accuracy of 82% on the test dataset.

Modeled a Support Vector Machine(SVM) in Python to classify whether a review represents a positive or negative sentiment.