I build practical software systems with a focus on local-first workflows, automation, native desktop apps, and reviewer-ready engineering documentation.
My current work is centered on Python tooling, SwiftUI macOS apps, agent workflow infrastructure, and study/project systems that are easy to run, inspect, and verify.
| Local-first | Tools that keep private inputs in the user's workspace and make state inspectable. |
| Deterministic | Rules, tests, and reproducible commands before AI or heuristic layers. |
| Reviewer-ready | Repositories with clear setup, verification, failure modes, and public/private boundaries. |
| Project | Stack | Why it matters |
|---|---|---|
| AudioConverter | Swift, SwiftUI, FFmpeg | Native macOS batch audio conversion and ordered merge app with deterministic UI-test seams and FFmpeg execution coverage. |
| Study OS | Python | Filesystem-first study planning CLI that turns course scope, source references, and recall results into deterministic learning packets. |
| SAP Clean Core Advisor | Python, Streamlit | SAP legacy landscape assessment tool that keeps scoring, TCO, risk, recommendations, and evidence traceable. |
| Codex Skills | Python, Codex skills | Reusable Codex skill library focused on making repositories easier to review, run, and reuse without inflated claims. |
| Territory War | C++20, SFML, CMake | Playable real-time strategy prototype with simulation rules separated from rendering and regression-tested game behavior. |
Primary project stack: Python, Swift, SwiftUI, GitHub Actions, and shell scripting.
Project experience: C++20, Java, TypeScript, CMake, Streamlit, and FFmpeg.
Workflow: AI-assisted development with Codex, backed by local verification, readable setup docs, and reproducible test commands.
- Prefer deterministic behavior before AI or heuristic layers.
- Keep private input data out of public repositories.
- Document commands that were actually run, not commands that should have worked.
- Treat setup, tests, and failure modes as part of the product surface.
- Keep automation useful, inspectable, and reversible.
I am refining a set of portfolio projects around local automation, native desktop UX, deterministic study systems, and reusable agent workflows. The goal is simple: repositories that a reviewer can clone, run, test, and understand without guesswork.