For the last 9 years I've built and shipped production software — the unglamorous discipline of making things that keep working when you're not looking at them. That foundation is the whole point. It's what separates "made a chatbot" from "ships reliable AI systems."

Across those years: full-stack web (React, Next.js, Vue, Angular, React Native, Node), cloud and serverless on AWS with infrastructure-as-code, Docker and CI/CD, and a run of SaaS products shipped end to end. The AI is the sharp edge on top of all of it.

These days that foundation goes into AI. I've built a custom MCP server deployed as a hosted, secured HTTPS service — with auth, scoping, rate-limiting and audit logging — so AI agents can safely use real tools over the network. I've built a multi-agent orchestration framework that routes work across specialised agents with review gates and clean hand-offs. And I've built LLM automations that run unattended and give people back hours every week.

Let's be precise about the label

I'm an AI application / agent engineer. My work is the application layer: agents, RAG, MCP, tool-calling, orchestration and evals — building real systems on top of foundation models.

I am not an ML or model-training engineer. I don't train or fine-tune base models, and I'll tell you that up front. Being honest about the camp I'm in sets the right expectations and means you get exactly the engineer the job needs.