(This page was last updated in Nov 2022)
Note: Click on any of the project names in order to view a short description of the project. For many of the projects, I link to a blog post with more details.
Improving the Strength of Human-Like Models in Chess [Blog] [Workshop Paper]
In this paper, we extend the concept of curriculum learning to improve the task performance strength of human-like models, specifically chess. We investigate the effects of choosing curricula and teachers, and find that weaker human-like models lead to better overall performance than strong non-human-like models. This paper has been submitted and is under review.
Ethical Impact of AI Predictions [Blog] [Paper]
In this paper, we investigate the ethical impact of predictions when assisting a human in the context of a Human-AI team. We ran a set of human study experiments on Amazon MTurk, and found that predictions strongly impact human prior ethical preferences, and the effect is increased when the prediction is sourced from an AI, as opposed to a human expert. This paper was published at AIES 2022.
In this paper, I built a conditional logit model to better understand the formation of a social network in the context of Online Health communities. I used this work in my Honor’s Thesis, allowing me to graduate Magna Cum Laude. In addition, this work has been accepted to CSCW 2020.
In this paper, I built a keyword classifier to determine the health journey type taken by the user, and this keyword classifier was used to validate the results of a machine learning classifier which had already been built. This work was published at ICWSM 2020.
Spring and Fall 2018
In this paper, I built a machine learning classifier to automatically determine if a patient has passed away based on the content of their blog posts. This work was published at CSCW 2017.
An Autograded Exercise Book for Learning Python [Blog]
This project is a companion course to py4e.com which allows any student to run interactive exercises in Jupyter Notebook while learning Python.
Predicting Surgery Durations [Report]
Learning Chess Biases [Blog]
This was my second project at SIFT. I evaluated the robustness of Bayesian network models, and extended the functionality of the SUNNY algorithm to compute conditional probabilities in real-time.
Spring and Summer 2020
This was my main project at SIFT. I implemented a Monte Carlo sampling module to forecast potential conflicts between friendly agents in a domain lacking communication. I also created a workflow for combining the output of multiple HOMER modules into a single Bayesian network for decision making.
Fall 2019 to Summer 2020
Measuring Sentence Structure Complexity
Automating AWS Deployments
During my internship at Amazon, I was tasked with automating time-consuming steps needed in order to stage update deployments for the Elastic Load Balancing tool of AWS. Thanks to my efforts, the time to deployment was reduced from several days of manual, error-prone work to only a few hours.
Seeing like a Bike [Blog]
During Summer 2018, I joined the Civic Data Science REU at Georgia Tech. The goal of this project was to use mobile environmental sensors to model the experiences of urban cyclists, and use this data to design better cycling infrastructure. An abstract of our work was accepted at the 2019 CARTEEH Transportation, Air Quality, and Health Symposium, and I gave a talk at the conference in Austin.
Automating PDF Invoice Extraction
During my internship at API Outsourcing, my work centered around automatically extracting data from invoices and remittances in PDFs. Due to the unstructured format of PDFs, it is very difficult to extract the data from these documents automatically, and most of these tasks needed to be done manually – a huge bottleneck when you need to parse tens of thousands of documents every week. I was able to implement a new method for extracting data which improved efficiency of automated extraction by 14%.