Zoe-Alanah Robert

The widespread misuse and addiction to opioids - which include prescription pain relievers, heroin, and synthetic opioids like Fentanyl - has become a national crisis in the US. This ‘Opioid Epidemic’ has far reaching side effects, in everything from public health to economic and social well being. A number of government and private institutions have publicly available historical data which, given the right data analysis tools, can help us understand which areas of the nation have the highest… Show more

The widespread misuse and addiction to opioids - which include prescription pain relievers, heroin, and synthetic opioids like Fentanyl - has become a national crisis in the US. This ‘Opioid Epidemic’ has far reaching side effects, in everything from public health to economic and social well being. A number of government and private institutions have publicly available historical data which, given the right data analysis tools, can help us understand which areas of the nation have the highest risk of a local overdose epidemic. As many opioid addicts ultimately turn to illegal forms of the drug, there is a known marketplace on social media. Additionally, pop culture has effectively popularized the use and abuse of opioids. These, along with existing research, suggest that Twitter data might hold some predictive power in assessing community risk of an opioid overdose outbreak. This project analyzes the use of Twitter data and historical public health data with neural networks to predict future community risk. Show less

Millions of people worldwide are absent from their coun- try’s census. Accurate, current, and granular population metrics are critical to improving government allocation of resources, measuring disease control, responding to natural disasters, and to studying any aspect of human life in these communities. Satellite imagery can provide sufficient infor- mation to build a population map without the cost and time of a government census. We present a CNN architecture which efficiently and… Show more

Millions of people worldwide are absent from their coun- try’s census. Accurate, current, and granular population metrics are critical to improving government allocation of resources, measuring disease control, responding to natural disasters, and to studying any aspect of human life in these communities. Satellite imagery can provide sufficient infor- mation to build a population map without the cost and time of a government census. We present a CNN architecture which efficiently and effectively combines satellite imagery inputs from multiple sources in order to accurately predict population density of a region. In this project, we focus on rural villages in India and use the 2011 SECC census as label data. We explore the effectiveness of multiple CNN architectures and various satellite input sources. We are ul- timately able to achieve an R2 of 0.9889, higher than all previous work performed on other countries. Show less

• Worked with a team of 3 other engineers over 6 months to iteratively ideate, program, and design a mobile application that launched on the App Store
• Spearheaded development on the React/Node/SQL/Ethereum backend and assisted with programming the React Native mobile app
• Pitched to leading tech companies and VCs, winning 1st place for best senior project from Garage Technology Ventures