- Lead extensive feature engineering, data pipeline development, and monitoring to create predictive telematics features abstracted across arbitrary sensor sources (mobile, OBD-II device, OEM) for use in behavioral driving risk models for >= 100k customers. Worked in close collaboration with telematics engineering team to design feature code to operate efficiently in both streaming and during backfill of 10s of TB of historical telematics data (Python / Airflow / Jenkins / SQL)- Developed several models}end-to-end across the Car product from inception through deployment and impact monitoring, working closely with stakeholders across the insurance, underwriting, actuarial, and engineering teams. Including regression, classification, and survival models built with linear or tree-based approaches, with a focus on impact, explainability, and maintainability over complexity.- Produced internal vehicle attribute risk model trained on over a billion rows of 3rd party loss data that reached a x2 improvement in risk segmentation and supporting more rapid future iteration (Databricks / PySpark) - Data science model and machine learning infrastructure improvements to integrate large volume of telematics data (billions of miles of driving with 1 Hz sensor sampling frequency) and telematics-based models into internal ML tooling, working closely with MLE team (Python/SQL/Airflow/optimization)- Creator/maintainer of internal package used across DS team (about 25 members) to drive AI/ML model governance with "model cards" creating transparency and maintainability (Python env management/MLFlow)- Consistently generalized novel project code for use by broader DS, DE, and MLE teams (about 35 members) via shared packages, with typing, documentation, and unit tests (Pydantic / package management / CICD)

- Consulting (5 hrs/wk) for an app-based agg-tech startup leveraging low-power, LoRaWAN sensors to monitor and control components of rural water systems (tanks, wells, pumps, etc.). Worked to define potential of high frequency time-series data in a space with no prior monitoring solution.- Lead R\&D and implementation of data-driven machine learning algorithms used for intelligent, automated pumping and water use anomaly detection. Supervised engineers implementing solutions. - Lead prioritization of decisions on everything data, including collection, storage, analysis, and end-user dissemination. Participated in short and long term project prioritization for company.

- Data Science lead of Metromile's core behavioral driving model, owning feature engineering, stakeholder interactions across multiple departments, ETL and model deployment, and documentation for legal filing. Leveraged high frequency telematics data covering over 1.25 billion miles driven to develop a new version of model, leading to a 35% improvement as measured by Gini score.- Owned the end-to-end model life-cycle from EDA through to production deployment for three models, both internal and customer facing, yielding significant improvement in each over existing models (Python/Sagemaker/Databricks/Airflow). -Developed shared utilities for data science team for model training and model monitoring, along with leading the development of a standardized, scale-able workflow for efficient model building and deployment (Python/Databricks/MLFlow/Sagemaker/Git)- Worked to retire existing legacy modelling infrastructure in favor of a more modern machine learning tech stack. - Maintained careful attention to data integrity and understanding core data leading to the uncovering and fixing of long standing bugs and issues in existing data ingestion pipelines and production models.

- Designed TrailMapper routing optimization algorithm to automatically generate interesting hiking and trail running routes that match user-provided constraints (Python) (git.io/JUbX4).- Data sourced from full OpenStreetMap database to enable route optimization worldwide. - Deployed interactive Plan-It Trails web-app (Python/Flask/JS/HTML) on AWS (git.io/JUbXw).- Moved quickly to design, develop, and deploy Plan-It Trails in under 3 weeks, with no prior experience working with geospatial data, routing algorithms, or developing a web applcation.

- Organized and led collaboration across several research institutions to develop actionable analysis plans for >100 TB of volumetric data and secured computational resources (3.5 million CPU hours).- Optimized agent-based model of individual stars in galaxy scale simulations in C++ to operate simultaneously on hundreds of galaxies for a statistically representative sampling of outcomes.-Built a first-of-its-kind rapid likelihood estimation and validation framework in C++ and Python for models of stellar nucleosynthesis leveraging observations of > 100k stars.

- Developed the first (and only) agent-based model following individual stars across an entire galaxy in astrophysical hydrodynamics simulations in C++, scaled to run on >100’s of CPUs.- Implemented new features and functionality to large, open-source astrophysical hydrodynamics codes alongside $\sim$20 other developers and ~100 users across many institutions.- Contributed 54 pull-requests with over 40,000 lines of code (C / C++ / Fortran / Python / MPI) across three projects: Enzo: (git.io/JUPyj) ; Enzo-E/Cello: (git.io/JUPSf) ; Grackle: (git.io/JUPSJ).

Assist students academically by facilitating a deeper mastery of course material in a range of math and science courses.Guided students in developing better study habits, seeking an understanding, not just memorization, of course materials.Maintained reputation for conducting excellent study/tutoring sessions, in a one-on-one context, up to managing several students at once.

Conducting research for senior thesis under Dr. Tom Jones with the Minnesota Institute for Astrophysics and Minnesota Supercomputing InstituteStudying the evolution of weak magnetic fields in turbulent plasmas through computational simulations in the context of the intracluster medium

Conducted radio astronomy research under Dr. Lawrence Rudnick, studying faint radio halos in galaxy clusters with a novel technique.As part of research, secured grant funding through the University of Minnesota's Undergraduate Research Opportunities Program.From work, presented a poster at a national conference, and published a co-authored paper in a major scientific journal (see publications)

Coordinated and executed transition of textbook ownership by Pope John XXIII High School, to ownership by students; worked as liaison between the school and Follett Educational Services, and managed communication to parents.Updated school wide computing resources; improved wireless network and Smart Board infrastructure.

Conducted nuclear physics research under Dr. Ralf Rapp, through the National Science Foundation's competitive Research Experience for Undergraduates (REU) program.Work resulted in a poster presentation at a national conference, and a primary authored paper in a scientific journal (see publications).

Received funding to work under Dr. Pablo Yepes at Rice University in medical physics.Improved code written to provide accurate and quick calculations necessary to properly administer proton radiation therapy to cancer patients in collaboration with researchers at M.D. Anderson