Law and operations taught me messy systems. AI taught me how to build them.

I build applied AI systems from real workflows: agent memory, orchestration, automation, and production tooling.

My path into tech is not the standard one. I come from law, advocacy, operations, and management — not a classical computer-science route. That background shapes how I build: I care less about clean demos and more about what happens when state, people, deadlines, errors, and edge cases collide.

Systems that help agents keep useful context beyond one chat:

• journals
• retrieval
• replay
• correction
• behavior that can be tested

Multi-step systems where agents:

• plan
• execute
• review
• retry
• hand work back to a human when needed

The hard part is not the prompt.
The hard part is state, contracts, failure modes, and orchestration.

Automation around creative and operational workflows:

• video / shorts pipelines
• image generation
• bots
• dashboards
• backend APIs
• frontend control surfaces

I use AI coding agents heavily, but I don't treat them as magic.

The hard part is still judgment:

• clear boundaries
• migrations
• contracts
• logs
• tests
• rollback paths
• knowing when the output is wrong

Python · TypeScript · FastAPI · React · PostgreSQL · Redis · Docker · LangGraph · n8n · LLM agents