Joseph Tully Email & Phone Number

Seattle, Wa, Us

Renton, Washington, United States

Teaching introductory CS courses for transfer students, involved in committees related to utilizing AI for education.

Served as deep learning researcher and advisor to a tech startup targeting automated design-to-code webpages.Created DL training environment and Torch codebase from scratch, utilizing AWS services (SageMaker, EC2, S3).Investigated potential solutions and rapidly developed prototypes for different tasks (object detection, classification, segmentation) by reviewing relevant literature, implementing model architecture and training, and adding experimental features to the existing app.

Bellingham, Washington, United States

Team member and co-author on a deep learning research project, in collaboration with Pacific Northwest National Laboratory.Utilized Deep Learning to improve on existing techniques to predict energy use in building climate control systems.Trained, evaluated, and tuned (using Weights & Biases) dozens of Torch models based on Google’s WaveNet architecture, created visuals and reports for PNNL collaborators, and modified training architecture based on results / feedback.Our work resulted in a publication in the May 2023 edition of the academic journal Energy and Buildings.

Bellingham, Washington

Lead 2hr weekly lab sessions for a variety of CS courses and levels: (Computer Programming I, Data Structures, Computer Systems I, Automata & Formal Language Theory)Introduce assignment with a short lecture, assess student assignments, and correspond with professor.Hold office hours to help students understand their course material and lab work one-on-one

Mentor computer science students weekly as a University employee. Help students with their programming assignments, assist in debugging code, and explaining concepts from their coursework.

Project Sharkzor - Worked on a team researching few-shot transfer learning to improve PNNL’s image classification product.Conducted experiments using various combinations of pre-trained CNN encoders (Resnet, Densenet), classifiers, and datasets to establish a baseline for work involving multi-CNN training.Implemented PyTorch GPU parallelization to allow for tuning multiple large CNNs on separate devices in parallel, increasing the training capabilities of the project

Master Of Science - Ms, Computer Science

Activities and Societies: Deep Learning Research Group, Whiteboard Coders

Bachelors, Computer Science