- Invented and open sourced a framework for multi-task, multi-modal, multi-dataset training for large ensembles of neural networks- Automated outfitting using multi-task neural networks- Object detection and segmentation for fashion use cases- Building and implementing GANs for industry use casesSpecialist in computer vision and deep neural networks. Dataset gathering, training, deployment. Work tasks involve building and deploying models for attribute modeling using multi-task ensembles combining custom CNNs and NLP methods. Also prototyping in house object detection and image segmentation tailored for e-commerce space

signal processing, object detection, and deep learning focused projects in classified space.

Currently working on counterfeit art detection using Siamese Convolutional neural networksI enjoy projects working with unstructured data in particular text and images to apply Natural Language Processing and Neural Networks (mostly work with Convolutional NNs). However I have been working to apply various machine learning techniques to public sector problems. Other tasks:-Automation of standardized reporting, data integrity maintenance, and mostly saving people time with scripts. -Developing statistical data driven KPIs, automating the analysis of KPIs (resulting in significant time savings for clients), easier articulation of progress made based on newly developed KPIs, deployed models including LSTMs for forecasting-Avant Guard: Presented to 500 members of Booz Allen senior leadership, Then built Siamese Convolutional Neural nets to detect counterfeit artwork.-Used NLP methods to identify dark web users-Led data science teams in 2017 and 2018 data science bowls as well as the Kaggle TSA passenger screening competition-Segmentation problems using Unet/LnikNet architectures -Object detection and facial recognition

Social Media and Communications for the McCourt School of Public Policy

Anti-Human Trafficking research project focused on Malaysia. Mined advertisements off of Malaysia's backpage.com and used data science to generate potential leads on human trafficking cases. Used text, phone numbers, time, and geographic spread to look at the complexity/level of organization required for different groups operating on backpage.com.Used a python based web crawler and a mixture of R, Python, and excel for data extraction and feature generation.