Multimodal Machine Learning for natural disaster prediction

- Contributor to Hugot, a tool making HuggingFace Transformer Pipelines available to Golang developers. - Implemented a new pipeline, adding support for Zero Shot Classification pipelines - Created single and multithreaded baseline tests for Hugot, comparing its performance to running transormer pipelines through Python

Using Machine Learning and Computer Vision to classify BMX tricks for the Australian National Team

Developed automated service to scrape data from Modeling and Simulation Confluence and Bitbucket pages, storing page information in MongoDB • Containerized application using Docker and deployed to Kubernetes cluster • Set up Cronjob to automatically run the service daily, updating the database • Replaces manual process of retrieving and storing data, saving hours of work on a regular basisCreated API function for a space architecture modeling tool to programmatically retrieve modified files from different git branches and move the files to a running Docker container • Programmatically runs shell commands to retrieve and move files • Prepping the tool no longer requires expert knowledge, with function automatically setting up the tool based on modified source files

- Worked with TROPICS team (Time-Resolved Observations of Precipitation structure and storm Intensity with a Constellation of Smallsats). Refactored data analysis and visualization scripts, greatly increasing accessibility and usability. Code interpolates geolocated brightness temperature data from the satellite’s passive microwave scanners to an ideal scan. Outputs latitudes, longitudes of interpolated ideal scans and creates a pcolormesh displayed on a world map. Implemented Naval Research Laboratory colormap and colorbar to best visualize data. Code will be used to monitor and analyze tropical cyclones in the the Pacific and Atlantic oceans. Supervised by Nick Zorn.- Also worked with the ATMS team (Advanced Technology Microwave Sounder) writing MATLAB code to direct the flow of data to appropriate readers for processing and visualization. Supervised by Vince Leslie.

Collaborated with fellow interns to design PPE using Solidworks for surgeons to reduce COVID transmission under Dr. Brian Wong. Edited multi-hour raw footage of rhinoplasty surgeries into concise 20 minute segments for teaching purposes.

Compiled databases of homemade face masks, PAPRs (personal air-purifying respirators), and ventilators, analyzing efficacy and production costs, under the supervision of Doctor Brian Wong. Database used by the Bridge Ventilator Consortium in book titled "Mechanical Ventilation Amid the COVID-19 Pandemic: A Guide for Physicians and Engineers." Co-author of chapter 23, "Alternatives to Conventional Non-Invasive Positive Pressure Ventilation Devices."