Agent architecture & ops
Scope boundaries, dispatch surfaces, human-in-the-loop approval, audit logs, and fleet patterns — so each agent has a name, a host, and a stop button.
Consulting
All consulting. One stack.
Engagements center on agents, context, and tokens — how your team designs, supervises, and pays for production AI.
Context & contract design
Durable agent context (think context.md, not chat history): what tools, paths, and credentials belong in scope; what must never enter the model.
Token & retrieval strategy
Context window budgets, tool-result hygiene, MCP and RAG boundaries — spend tokens on decisions and diffs, not repeated system prompts.
Harness implementation
Hands-on work on dispatch, runtime integration, tmux supervision, and logging — including JARVIS-style patterns where they fit your estate.
Workshops & speaking
Executive briefings, engineering deep dives, and classroom-style labs — AI-assisted development that ends in deployable systems, not toy demos.
Architecture review
Honest assessment of where you are thin: model choice, prompt soup, missing supervision, or infrastructure the agent should never have touched.
Hire
Book Matt by the hour.
Consulting is $125 USD per hour. Choose how many hours you want, pay securely with Stripe, then we schedule the session.