I help teams stop experimenting with AI and start using it.
Working AI systems for business teams — not developers. Built with guardrails, human approval gates, and documentation your team owns. Every claim below is documented in full on the portfolio site, including a production pipeline with live public stats.
What I Build
Agent pipelines for high-volume, quality-sensitive work: model routing by task value, hard guardrail gates, review queues — nothing irreversible without a human click.
Role-specific training and playbooks for non-technical teams — executives, marketing, operations — so adoption survives after I leave.
Discriminator-task benchmarking on your actual workload: which model, for which tasks, at what cost — evaluations that end in spending decisions.
Quick-start engagement: a deck & content production system — consistent, brand-locked presentations and documents on a prompt + rubric system your team keeps.
Receipts, Not Claims
How an Engagement Works
Workflows can run on the Claude subscription your team already pays for — no API keys, no per-token billing — or as a multi-model API build when volume demands it.