Bhavesh Suhagia

Designed a convolutional neural network for end-to-end driving using TensorFlow/Keras. Used optimization techniques such as regularization and augmentation techniques to generalize the network to drive on multiple tracks.

Build a computer vision pipeline to detect highway lane lines given real-world image. Used OpenCV library to perform image analysis techniques including Hough Transforms and Canny edge detection to complete the task. The pipeline was then tested on a real-world video sequence

Collected and preprocessed S&P 500 stock data from 2010 - 2015 using R library quandmod. Built various clustering methods such as k-means and hierarchical methods to group stocks into different buckets which represented the volatility of the stock index.

Analyzed Amazon text reviews and extracted relevant features using LDA method to discover latent structures and vectorize the raw text. These features were then used to build a sentiment classifier using Naive Bayes in Python. Finally, created a visualization dashboard to display results using Shiny (R).