I helped develop the back-end framework of a Gies College of Business data hub using Django to gather and organize research data. We leveraged AWS technologies to dynamically pull from various data sources and maintain up-to-date, flexible data to facilitate future research projects. I also contributed to the creation of a website UI to house our data hub using React and Figma to create intuitive and user-friendly front end.

I trained several computer vision models using Darknet and YOLOv3 to identify CAT vehicles and people, with the aim of strengthen safety measures in worksites. To improve the efficacy of CV models, I automated the process of synthetically generating training images using Python and Blender’s Python API, which allowed us to diversify training datasets. I also conducted a study to measure the impact of label variation on model efficacy and mAP scores using Amazon SageMaker and Docker containers, in order to better curate training data for future models.