Research & Projects

I am involved in research and personal projects focusing on the use of AI to solve real world problems. My work involves the use of AI/ML in realms from conservation to education.

Remote Sensing with ML: Stony Brook University Simons Research program

Worked with Dr. William Holt at Stony Brook to develop a novel system for detecting offsets in GPS time-series.
Developed sliding window algorithm to generate a dataset of GPS data labeled with offsets.
Tested a a number of different classifiers to develop a model with F1 score of .98 on both classes

Click the image above to see my abstract, published at the American Geophysical Union

Predicting Snow-Water-Equivalences with Neural Networks:
Research @ George Mason University

Worked with Dr. Ziheng Sun in Geoinformation Science Department to make Snow-Water-Equivalence (SWE) predictions publicly available with ML models.
Developed and tested a number of Neural Networks models to predict future SWE given a location.
Integrated models into SnowSource app to provide SWE data

Berryville Institue of Machine Learning

Worked with world renowned computer scientist and security expert Dr. Gary McGraw on his latest research venture of identifying security issues in ML algorithms at his new startup BIML.
Read through research and curated bibliography which has become a go to site for researchers interested in security of Machine and Deep Learning algorithms.

Save Our Seas (SOS)

Developed an edge-capable AI machine-learning model for marine garbage detection in Python on Linux.
Successfully trained using PyTorch a single-shot object detection model for detecting multiple types of garbage and fish within a single frame.
Built the model in the AWS cloud, which was then converted to ONNX and deployed onto an edge device.
Tested and evaluated metrics on the model for different dataset sizes and number of epochs to optimize the performance of the classifier

Drone Enabled Env. Patrol & Surveillance Edge AI System (DEEPSEAS)

Developed a prototype to reduce illegal, unreported, and unregulated fishing reaching the status of Conrad Innovator in the 2022 Conrad Innovation Challenge.
Led technical direction working with cross-functional marketing and product team across hardware and software to detect illegal fishing vessels on an edge device with a camera mounted on a buoy.
Designed the end-to-end technical architecture and its integration with hardware components that resulted in building the model and reference engine.
Developed the machine learning model and inference engine deployed to NVIDIA Jetson Nano 2GB on Ubuntu with a serial camera

Late-Graduation Risk Assessment Shenandoah Valley Comp Sci Regional Partnership–George Mason Univ./VDOE

Developed a Random Forest Decision Tree Model and a Logistic Regression model to predict whether a student, given middle school grades, attendance, and demographic will graduate high school on time.
Gathered and preprocessed historical data of students in the Winchester Public Schools system.
Ran analytics to determine the efficacy of models using an iterative process to continually tune them
Achieved a model that can be used by schools to better target students that need additional assistance