I direct AI coding agents to build software, define what successful work must demonstrate, and review outcomes before accepting them. My focus is AI orchestration, workflow reliability, and verification.
Seeking remote AI implementation, automation, or agent operations opportunities.
Woozy VTuber — a Blender-built VTuber character with real-time face tracking, OBS integration, and a documented v1 → v2 rebuild. The project I know inside out: I set the requirements, directed the build, and tested every iteration live on stream.
- Case study — what changed from v1 to v2 and why
- Live on Twitch: twitch.tv/woozyfilms
AI Orchestration & Verification Portfolio
Four case studies from my own development environment:
- Verification system — a deterministic controller that checks repairs against a frozen contract.
- AI-worker supervision — catching an unrelated test rewrite and investigating sandbox-specific test failures.
- Python-Inspector (public source) — a desktop code-inspection tool with Docker test execution and documented validation.
- Evidence-durability repair — preserving verification records through reinstalls.
I set goals, scope, permission boundaries, and acceptance criteria. AI coding agents implement changes and run technical checks under that direction. My portfolio describes that division of work, the recorded outcomes, and the remaining limitations.
The public case studies include screenshots and dated results. Underlying implementation code and raw evidence for the Verification system, AI-worker supervision, and Evidence-durability repair case studies remain local or private. Python-Inspector is the exception: its full source, commit history, and test suite are public at github.com/hunter-terry/python-inspector.
Tools used in this work: Claude Code, OpenCode with OpenRouter models, Codex, Python, PowerShell, Docker, and Git.