Agentic AI · advisory · coaching · workshops · training
The model is the engine. I build the harness around it.
I help teams get real, reliable leverage from the AI they already pay for — by engineering the system around the model, not just the prompts. My deepest work is with developers on Claude Code; the same thinking serves product, ops, and any team. Hands-on, on your setup, in real time.
What I actually do
Close the gap between "we have Claude Code licenses" and "our team ships measurably faster with it."
Build the harness
Context, custom agents, skills, hooks and observability that turn a raw model into something a team can trust on real work.
Make it reliable
Guardrails, verification, evals, model-offloading and audit trails — agentic workflows you can actually deploy and see.
Run it without you
Loop engineering: scheduled, unattended agent runs (cron) that do real work while you're out of the loop — the harness is what makes them safe to trust.
Teach by showing
Live work on your setup, an honest before/after, and the lessons learned the hard way — including what not to do.
Two ways I teach it
Applied — your setup
Hands-on on your own Claude Code setup and daily workflow: build the harness, make it reliable, ship faster. For individuals and teams.
Foundational — how LLMs work
The mental model under the tools — tokens, context windows, inference, training, what an agent really is — so your team makes sharp calls, not cargo-culted ones.
New here? Start with how I work & how I earn your trust, or go straight to sessions & pricing.