Technologies: AWS, Azure, Caffe, Python

For this project, I

• Prototyped recommendation algorithms to suggest dishes based on user location and taste using Keras, Azure and AWS machine learning suites and achieved 0.7+ Spearman correlation


• Deployed the best scoring algorithm as a scalable prediction API to AWS using S3 and ECS


• Created a MongoDB database of dish names, descriptions and taste levels using data from Locu and Yummly APIs for 100K+ restaurants in the US and the UK

Technologies: C++, ARCGIS, Caffe, MATLAB, Python

For this project, I

• Developed a pixel-based computer vision algorithms to detect buildings in satellite images with 0.86+ AUC score using MATLAB for image preprocessing, C++ for image segmentation and Caffe for deep learning


• Created and open-sourced a ground-truth dataset of satellite images, structural footprints and heights of 40K+ buildings in 7 US cities and suburbs using ARCGIS, OpenStreetMap and LIDAR

Technologies: D3.js, Google Maps, Gephi, Leaflet.js, Python, R, Tableau

For this project, I

• Visualized a network of 1M+ travel origins and destinations searches by Expedia users using Gephi, Leaflet.js with Google Maps layers, and Tableau


• Performed user segmentation on 100K+ users through K-means clustering in R and extracted hotel choice insights based on date, travel distance and group size

Technologies: Google BigQuery, R, GDELT , Tableau

For this project, I

• Constructed and optimized SQL queries using Google BigQuery to extract data from the 2.5 TB+ global knowledge graph from Global Database of Events, Languages and Tones (GDELT)


• Created sentiment time series analysis of news coverage of the 2016 Elections and its outcome using Tableau visualizations

Technologies: Azure, D3.js, Python, R

For this project, I

• Created interactive visualizations using D3.js and Tableau for housing data in Ames, Iowa


• Built prototypes of prediction algorithms using Scikit-Learn through feature engineering, gradient boosting and ensemble learning and achieved 0.72+ cross validated R2

Technologies: Python, Mathematica

For this project, I

• Built a proof of concept circuit capable of detecting muon to electron (Mu2e) conversions currently used in Beyond Standard Model physics research at Fermilab


• Modeled chaotic dynamics of double pendulum using Python