AI-native backend and platform engineer in Austin, six years in. Most recently at Arbor, I built production LLM agents, the approval and audit layer they write to the database through, and the evals that decided when they could run unattended.

I'm currently experimenting with retrieval, dollar cost per run, and prompt-injection defenses for agents that write to production.

Selected work

• scriptorium. Claude Code skills for writers and serious readers, with precision proofreading and review that works inside the author's own frame.
• VibeCraft (Gauntlet AI, late 2025). An AI music video pipeline that analyzes a song for tempo, sections and lyrics, then assembles a beat-synced 1080p cut from generated video. A paired project, built in ten days.
• message-ai (Gauntlet AI, late 2025). A React Native messaging app with real-time chat and six AI features, including semantic search over embeddings.
• Marionette (Gauntlet AI, late 2025). A computer-use agent for macOS driven by natural language, my team's capstone. There's no public code, but my demo segment shows it working.
• Haven. An open-source secrets manager for small teams, which I co-created. The case study explains why and how we built it.

I'm looking for a mid to senior individual contributor role, in Austin or remote. Find me on LinkedIn, and ask me my favorite agentic coding patterns and skills.