I design and operate AI-assisted workflows that turn messy business processes into reliable systems. Most of that work is n8n: multi-step pipelines that pull from APIs, route work through LLM stages, write state back to a source of truth, and hold a human approval gate before anything goes out.
I build these against live client accounts, not test environments. That means most of what I actually know is about failure modes — batch updates that silently write only the first row, config changes that cascade into the wrong month across an entire calendar, model output that passes a rules check but is factually wrong. Catching those is the job.
Recent work includes a six-workflow content production system running a daily automated cadence across two brands in a regulated finance niche, an SEO audit-and-repair engine that pushes fixes through the WordPress API, and a lead research pipeline that enriches contacts and hands them to a CRM with tagging that triggers a multi-touch sequence.
Before automation I spent eight years in operations — B2B ordering, e-commerce support, and process control monitoring order data end to end through an SAP and ESB stack. That's where I learned to look for where data breaks between systems, which is the same instinct the automation work runs on.
Tools: n8n, Anthropic and OpenAI APIs, Google Workspace APIs, WordPress REST API, GoHighLevel, JavaScript, JSON, webhooks.
Case study and portfolio: -----------
