➤ Optimized an existing legacy production feature, boosting prediction accuracy by 50% and reducing training time by up to 100x. Incorporated a comprehensive test suite using python's pytest and unittest libraries ➤ Created an NLP tool for sensitive information sanitization in call center dialogues. Automate deployment via an asynchronous SageMaker endpoint with autoscaling. Integral to processing 1M+ daily transcripts in Calabrio's data pipeline ➤ Orchestrated an end-to-end build and deploy process using ADO pipelines for library artifact creation, Docker container building, AWS ECR integration, SageMaker deployment, lambda entry points, autoscaling, and validation tests for beta cloud features. ➤ Engineered proprietary algorithm detecting voice agitation in audio, in the context of call centers, using clustering, custom speech-to-vector encoders, and fine-tuning ResNest using spectrograms. ➤ Fine-tuned a neural network (BERT & DistilBERT) to classify a transcribed conversation as pertaining to a specific user-defined phrase category. ➤ Training a sentiment analysis model using a transformer neural network (ELECTRA) to predict sentiment mixture probabilities.

➤ Trained a neural network (CNN) to encode speech to vectors for automatically identifying when an agent is speaking during a call. ➤ Developed an algorithm to separate mono-channel audio into two speaker homogenous channels -- outperforming third party vendors being explored by Calabrio ➤ Designed production machine learning deployment architecture using Docker broker-worker pairs integrating with the core Calabrio platform for training and inference.

Expert on close stellar binary systems and the generation of magnetic fields in stars and their effects any attending exoplanets. ➤ Experience in handling large data sets, used SQL queries to isolate ∼ 2000 objects of interest from >100 million objects [Python (pandas), SQL] ➤ Designed new innovative metrics using ensemble classification algorithms for logistical binary classification [Python (scikit-learn)] ➤ Wrote an algorithm to classify objects from a grid of 57x372 continuous parameters using chi-squared minimization techniques [IDL] ➤ Used Fourier analysis, signal processing, and least-squares spectral analysis to determine periodicity in time-series data [Python (scipy)] ➤ Developed software for reduction of spectroscopic data for the DeVeny spectrograph at the 4.3-meter Discovery Channel Telescope [IDL] ➤ Refined technical writing skills in writing peer-reviewed journal articles and several successful telescope observing proposals

➤ Public Outreach – I volunteer at the Boston University Astronomy Open Night. We provide the public to weekly (weather permitting) access to small telescopes on the roof of the Astronomy building ➤ Undergraduate Mentor – Designed and oversaw research projects for undergraduate students in order to provide skills in independent research, data reduction, and data analysis necessary for pursuing a graduate school in astronomy ➤ Upward Bound – Mentored underrepresented local high school students in projects using real data addressing cutting-edge science questions ➤ RISE – Introduced promising high school students to tools required fordata analysis and reduction, data visualization, and critical thinking re- quired in independent research

➤ Discussion leader for Astronomy 105, Cosmology. I held four one hour mandatory sections a week and I had complete freedom in choosing material. Students were asked to find interesting popular science articles relevant to the coursework and come to class prepared to discuss with their peers. I would help facilitate and guide discussion. ➤ Lab Instructor for Astronomy 202 - Introduction to Astronomy for Majors. I was in charge of running four three-hour sessions a week. These sessions including lab experiments using light and optics in astronomical contexts. There was also a night lab portion where students were asked to design their own astronomical research project using the 10" telescope on the roof. They were introduced to data acquisition, data reduction, and technical lab write-ups.

Developed a web app (three week project) that tests whether users can distinguish between real political candidate quotes and quotes generated by a bot. Website: http://www.polibot.us // code: github.com/dylanpmorgan/polibot ➤ Parsed, processed, and stored transcript and speech data from http://www.presidency.usb.edu using BeautifulSoup, Python, and PostgreSQL ➤ Implemented a self updating third-order Markov chain with Part of Speech tagging that generates grammatically correct sentences in the speaking style of the political candidate ➤ Stored user questions and responses into PostgreSQL database for real-time and future assessment of bot performance and user behaviors ➤ Deployed an interactive front end using Flask and Bootstrap deployed on Amazon Web Services

My research focused on using periodic stars as tools for probing structure in the outer Milky Way Galaxy. Analysis required Fourier analysis of time-domain data to determine periodicity of variable stars. As well as 3-dimensional data visualization to determine structure in the Milky Way.