Joel Aftreth

- Deployed on AWS EC2 a Fastapi app with microservices architecture using Docker for predicting NBA shot accuracy - Includes a PostgreSQL database, OAuth2, ETL pipeline using GitHub Actions, DagsHub and MLFlow, React frontend and Prometheus/Grafana for monitoring (https://github.com/jja4/nba_mlops) - My app was selected as the prototype MLOps app for new students to learn from

Developed a machine learning model for early classification of Alzheimer's disease using EEG signal processing features in Matlab/Python, leading to a start-up valued over $1M Saved months of manual work by creating a web-scraping tool using Scrapy and Natural Language Processing (NLP) to extract prices and emails from web search results Created a C# .NET MAUI SDK for mobile application development, resulting in a 40% reduction in development time for new mobile applications

- Developed a computer vision approach to predict ion temperatures within the W7-X stellarator using convolutional neural networks (CNNs) and TensorFlow - Reduced prediction time by a factor of 100,000, with 92% relative accuracy, fast enough for near real-time control

- Developed a computer vision/convolutional neural network (CNN)-based brain-computer interface (BCI) for motor imagery classification using PyTorch - Achieved a 90% classification accuracy, outperforming the state-of-the-art linear model (82%)

Machine Learning and Physics models of the brain