As an Assistant Professor I do so much more than just teach. In this role I’ve broadened my practical and theoretical understanding of machine learning algorithms, I train and lead teams of student researchers, and I’ve honed my presentation and communication skills. I am the machine learning expert in the department, and I developed the machine learning course from scratch and oversee student research, masters’ theses, and capstone projects. I’ve developed a comprehensive code-base to tackle any kind of data science or machine learning problem ranging from information extraction, question answering, data visualization, machine vision, and predictive analytics. Additionally, I stay up to date with state-of-the-art deep learning algorithms, regularly publish my research, and collaborate with colleagues on administrative work and grant proposals.

My primary responsibilities included supporting lab research and conducting my own independent research. I managed graduate and undergraduate research, analyzed and summarized experimental results from larger collaborations, performed literature reviews for grants and publications, and organized and co-chaired the National NLP Clinical Challenges workshop and conference. My research focused on natural language processing for the clinical and biomedical domains. A few projects included developing deep learning systems for joint named entity recognition and relation extraction, and information extraction from social media for mental health applications, smoking cessation, and adverse drug event prediction.

I received my PhD in Computer Science with a focus on Natural Language Processing and Machine Learning. During my PhD, my research centered around semantics and knowledge discovery. Semantics is the study of how to represent meaning from language, and I studied methods to extract, represent, and manipulate meaning from text with the goal of discovering new knowledge from it. I developed methods to encode term meaning into vector representations (word representations), developed unsupervised methods to quantify the degree to which two terms are related (semantic relatedness), and developed methods to discover new knowledge implicit in text, but never explicitly stated (literature-based discovery).

I managed a software team that developed hyperspectral imaging products for the defense industry. My responsibilities included algorithm design and development, user interface design, quality assurance, team management, and meeting with clients for requirements gathering and results presentation. I worked with several interdisciplinary teams including mechanical and electrical engineers and the marketing department. Additionally, I worked in tandem with the marketing department to discover new applications of hyperspectral imaging technology. I developed algorithms for those applications and to explained technical concepts to potential customers.

I founded Optical Research Lab, LLC, a company focused on developing novel applications of hyperspectral imaging. As president and sole employee of the company, I took responsibility for all aspects of company success. This included the development of a comprehensive data analysis and data visualization software package, technical and market research, finding and meeting with clients, and completing contract work.

I led the development of a classification and analysis platform for predicting tobacco quality from hyperspectral images. I developed the system from concept to near line implementation including data collection, exploratory data analysis, classifier prototyping, hardware design, front and back-end software design and final implementation. The final product included a touch screen user interface, hardware interface, visualization, and real-time data analysis. The system allowed novice users with minimal training to outperform highly skilled experts. Additional projects involved exploring novel applications of hyperspectral imaging to industrial processes, resulting in several patents.

I created signal processing techniques and an algorithm for the automatic classification of avian flight calls. This was challenging due to the variation inherently present in biological acoustics. The large dataset and complexity of the algorithm required extensive parallelization on local and off-site computing clusters.