I'm a software engineer and marathon runner building local-first data tools and evidence-driven AI collaboration workflows.
I am the primary maintainer of a local-first Python CLI that turns Garmin Account Data Export into deterministic, auditable datasets for reproducible analysis. Processing stays local, unknown values and unresolved relationships remain visible instead of being guessed, and Run-All produces QA, audit, and provenance alongside normalized data.
The project is designed to make data quality boundaries reviewable before downstream analysis. A tracked synthetic workflow lets people evaluate the process without a Garmin account or real personal data.
Garmin Running Data Normalizer repository · Garmin Running Data Normalizer on PyPI · Synthetic Product Quick Start
A provider-neutral operating standard and reusable assets for structured AI-assisted collaboration projects.
I use ChatGPT and Codex as collaborators for research, implementation, testing, and review. Human review remains the decision boundary: product choices, evidence limits, and release authority stay under human control, while reproducible artifacts make the work inspectable.
Development notes and projects on Digital Hub · Technical articles on Zenn · Development Instagram · Development Threads
Running is supporting context for my work: I train for and run full marathons.