AI evaluationApplied AI engineering
At Snorkel AI, I create terminal-based benchmarks for coding, physics, mathematics, and geoscience. My software projects cover browser automation, computer vision, and full-stack AI products.
Available for AI engineering roles
Selected work
Six public projects with links to their source code.
Computational geometry
A browser-based CAD application for woodworking. It stores an editable feature history, rebuilds geometry deterministically, and exports shop drawings in SVG, PDF, and DXF formats.
Geospatial ML
A U-Net segments roofs in satellite images selected by address. The FastAPI and React application supports single-property, batch, and side-by-side comparison workflows.
Fintech AI
A portfolio assistant that calls typed analysis tools and checks its responses with confidence scores and validation rules. The repository includes a React interface and a FastAPI service.
Research automation
A research agent for Kindle Direct Publishing. It searches for niches, compares competing titles, drafts listing copy, and scores demand and competition with configured market data.
Browser automation
A CLI that turns natural-language test instructions into Playwright actions. It records sessions, writes Playwright tests, and tries coordinates when a selector fails.
Education
A tutoring application that responds with guiding questions. It records sessions, displays learning analytics, and supports real-time chat in the React client.
Experience
Snorkel AI
I create terminal-based evaluation harnesses and containerized benchmark environments for Terminal Bench and Reasoning Bench. I have authored more than 50 tasks across coding, physics, mathematics, geoscience, and data analysis.
I write reference implementations, golden answers, rubrics, graders, and deterministic checks. As a reviewer, I have found more than 100 reasoning, validation, and reproducibility defects.
Gauntlet AI
I completed a 10-week engineering program based on weekly product sprints. I built Ghostfolio Agent with FastAPI, LangGraph, React, Docker, evaluation cases, and LangSmith traces.
Hack the Gap
I built React, Node.js, and Express features and REST API integrations for internal and client-facing products. I added input validation, structured error handling, tests, accessible interfaces, and documentation for components and integrations.
University of Texas at Austin
I built Python and Jupyter pipelines for terabyte-scale 3D seismic data from Axial Seamount. I used them for data cleaning, quality control, anomaly detection, visualization, and seismic interpretation.
Technical focus
I use typed tool inputs, evaluation cases, confidence thresholds, and trace data to find incorrect or unsupported agent responses.
I build Python or TypeScript services and connect them to React interfaces. Tests and deployment configuration live with the application code.
I use maps, satellite imagery, 3D geometry, and numerical analysis in Roof Segmentation, Parametric Modeller, and other projects.
Background
I earned a PhD in Geophysics from the University of Texas at Austin, an MS in Geological Sciences from East Carolina University, and a BS in Geology from Gustavus Adolphus College.
Read my résuméCredentials
Environmental Geology and Hydrogeology certificate from East Carolina University. IBM Full Stack Software Developer Professional Certificate.
Teaching
I taught mathematics, physics, chemistry, and ACT/SAT preparation to more than 20 students. Their average ACT score increased by five points.
AECOM
I supported drilling oversight for more than 25 wells and borings and prepared groundwater, geospatial, and hydrogeologic analyses for technical reports.
Contact