Dylan Rodriquez

Built an end to end ML experiment orchestration, management, and CI/CD system utilizing technologies across Git, Docker, AWS, MLflow, Flask, and Kubernetes that is compatible with most machine and deep learning packages. Optimized data cleaning and engineering pipelines to reduce compute time to 1.67% of the original runtime (a reduction of 44.25 minutes). Reduced memory and compute requirements to 8GB with dual core CPUs in order to train biomedical datasets while reducing infrastructure costs.

Extended a model creation metalanguage for programmatic generation of models, feature extraction, and transfer learning. Defined problems from gastroenterology and inflammatory bowel disease into machine learning projects. Led annotation strategy and QC within the ML team and onboarded multiple new ML engineers.

Led the development of a big data analysis framework in Python utilizing Spark SQL and Nvidia Rapids identifying disease and diagnostic subtypes. Developed scalable Python and Apache Spark programs for processing "big data" and automated testing of machine learning models. Designed methods incorporating generative adversarial neural networks to preserve the privacy of data protected under federal (US) HIPAA regulations.

Instructor for object oriented programming in python with an average of 69 students in Fall/Spring and 10 in Summer with a focus on the fundamentals of programming and basic graphics implemented with the python standard library. Created online content for instruction to ensure continuity of classes across the face-to-face to online transition in 2020 Managed and trained undergraduate teaching assistants

Advised by Xavier Tricoche. Investigated the use of tensor field visualization for diffusion weighted imaging in assisting physician decisions.

Research programmer and analyst for clinical trials in Inflammatory Bowel Disease. Programmed automated data collection, analysis, and tracking of study compliance using consumer biometric sensors and mobile apps. Led and managed research projects; all of which resulted in publication with one changing US cancer screening recommendations. Developed mobile applications to provide real-time inflammation estimates of endoscopic images viewed through the phone camera.

Research analyst and programmer for clinical trials and academic research projects. Analyzed access to healthcare in the United States, localizing geographic and demographic areas in need. Engineered datasets using SQL and Python, reducing the need for full retrospective chart review. Collaborated with senior research personnel to propose studies and design study methodology. Managed the pilot study for fecal transplant (FMT) in ulcerative colitis and trained nurses in FMT treatment of C. difficile.

Managed investigator and industry sponsored studies. Researched patient perception of FMT (fecal microbiota transplantation) for a later pilot study of FMT in Ulcerative Colitis. Programmed statistical analyses for multiple studies.

Thesis: Evaluation of Archetypal Analysis and Manifold Learning for Phenotyping of Acute Kidney Injury. Advisor: Ananth Grama Co-chair: Mohammed Adibuzzaman Advisory committee member: Petros Drineas