Implemented a Mask R-CNN model using Meta AI’s Detectron2 platform to detect boulders from lunar imagesAssimilated various datasets to validate the model against traditionally used methods in the fieldPresented results at national science conferences, focusing on communicating ML aspects to non-expertsPublished findings in peer-reviewed journal and made the associated code open source on GitHubWon over $10k in university grants to fund the research projects

Coded a CNN-based pipeline to detect collapsed buildings from satellite images and building footprintsOptimized image preprocessing script to cut runtime by 75% and fix a memory leakPresented final product to company leaders for implementation in client-service workflow

Created a regression model pipeline using TensorFlow & Keras to obtain ice thickness data from radar imagesDeveloped a deep learning image classification model to label over 26k images based on scope type