Applied AI / forward-deployed engineer in San Diego.
I build AI agents, automations and business systems.
My work connects business workflows with software that can be run, reviewed and maintained.
Client work | webmoz.ai | LinkedIn
Nine runnable agent patterns: a fixed workflow with sub-agents, a router and orchestrator, agents as tools, context management, RAG with evals, a REST API exposed as MCP tools, human approval, event-driven jobs, and an evaluator retry loop.
- Each pattern is its own folder, with the implementation, a runnable demo, and tests.
- An invoice investigation service, CLI, and API run those same patterns as one workflow.
- Python, LangGraph, FastAPI, SQLite, and MCP. Runs locally without API credentials.
Read the architecture | Open the workflow code | Run it locally
Instructions and templates for AI coding and creative tools. These are separate from the application projects above.
| Skill | Purpose |
|---|---|
| Plan to execution plan | Break detailed specifications into bounded implementation steps |
| Challenge plans | Review a plan, spec, or diff before anyone builds it |
| UI and backend alignment | Walk a live journey and match each screen to logs and stored data |
| Webflow to clean code | Move exported pages to self-contained HTML, CSS and JavaScript |
| Directing motion films | Direct motion graphics with briefs, references and frame reviews |
| Video script and storyboard | Script a product video around proof, then hand off scene cards |
| Talking-head clip review | Pick the best takes locally and assemble only what a person approved |
| Deploy dash normalizer | Apply a text convention to generated files before deployment; includes a standalone Node.js script |