PROJECTS

Object Detection Using Google Open AI Images

Transfer-learned and trained a convolutional neural network, YOLO using Keras and Tensorflow to identify 43 new objects such as guitar, beer, train, bus, woman etc. within images with. Used ~300K images to train the network on Google Cloud Platform (GCP) using NVIDIA Tesla K80 GPU for ~ 10 hours

Property Recommendations for HomeAway's Customers

Scraped ~ 8000 customer reviews for HomeAway’s (vacation rental marketplace) properties in Austin using Selenium in Python and grouped the properties into 4 different categories using Latent Dirichlet Allocation (LDA) model. Recommended category specific properties using Cosine Similarity between customer reviews and property descriptions

Non-Profit Crowdfunding Recommendation System

Developed collaborative filtering algorithm using Singular Value Decomposition (SVD) and content-based recommendation system using TF-IDF scores to suggest new projects to existing donors on DonorsChoose.org ( a crowdfunding platform connecting benefactors to schools)