Self Projects

Moodify: Mood-based Music Recommendation System

Implemented a feed forward neural network (FFNN) from scratch and a convolutional neural network (CNN) on MNIST data set achieving an accuracy over 97% after rigorous hyper-parameter tuning. Studied the principles of regression, stochastic and mini batch gradient descent, regularization and batch normalization and got acquainted with feature engineering and performance metrics. Designed the frontend of a web app (containing the CNN model) and deployed it using Streamlit.

Link: Code

Winter in Data Science: Autonomous Driving

A bootcamp in data science, followed by a project on Autonomous Driving - Object Detection and Driver Behaviour Analysis, was organised by the Analytics Club of IIT Bombay in the Winter of 2022.

I performed a comparative analysis of algorithms to give the best results by employing various ML architectures such as SVM, Naive Bayes classifier, and ensemble learning using the scikit-learn library. Comprehended the YOLO implementation in detail after studying different deep neural networks. Used various statistical techniques and visualization formats like line, box, violin, bar, scatter plots, histograms and correlation heatmaps in order to predict driver behaviour in different terrains.

Link: Code