Production AI and SvelteKit apps your team can own.
I'm Scott Spence, a product engineer and Svelte consultant in Kent. I build workflow products and agent systems, plus the checks that stop coding agents eroding your architecture. The checks are code, not instructions, so they keep working after I've gone.
Svelte Ambassador and co-founder of Svelte Society London. Previously I led two engineering teams building AI customer-service software. My MCP and coding-agent tools are open source, and I write up what I learn at scottspence.com.
What I do
SvelteKit products
Workflow apps that ship and stay supportable.
Greenfield SvelteKit builds from clickable prototype to production. Document intake, review and audit flows, clear service boundaries, and a codebase your team can pick up.
Production AI systems
Agents you can see inside.
Agent runtimes with tools, MCP servers and sandboxed execution. Evals, traces and review steps go in from the start, so you can tell when something has gone wrong and why.
Coding-agent guardrails
Checks that say no before a shortcut lands.
An architecture document doesn't stop an agent putting code in the wrong layer. A failing check does. I write boundary, data-ownership and release checks that run in CI and tell the agent where the code belongs.
Cloud and production support
Deploys, and finding out why they broke.
Deploying and debugging on AWS and Azure: containers, Postgres, storage, email, secrets, Terraform and DNS. When production misbehaves, I work from the logs and traces to the cause.
Engineering leadership
Hands-on, then a clean handover.
Contract tech leadership, code review, pairing and workshops. I've led two product teams at once, and I help teams use coding agents without lowering their standards.
Selected work
All workReinsurance underwriting platform: prototype to production in six weeks
A clickable prototype with no backend became an underwriting platform running in the client’s private Azure network about six weeks after the first commit.
Reinsurance
6 weeks From the first commit to running in the client’s private Azure network.
Claims platform for deposit-replacement insurance in five weeks
A spec in August became a claims platform in production five weeks later, built on a live platform I had been releasing since March.
UK rental deposit-replacement insurance
5 weeks From the written spec to the claims platform in production.
Telecoms sign-up app: prototype to production in under two weeks
A click-through prototype became a live multi-brand SIM sign-up app in about ten days. It has taken real customer orders since July.
Telecoms reseller
10 days From the click-through prototype to production on AWS.
Recent writing
All posts- The best architecture decision makes the next big change boring After handover, a client’s team made a major capacity change to an underwriting platform in one week, with no rebuild. It was a rules and database change because of one early decision.
- The client gets what they bought
- Guardrails that outlast the engineer
- The production failures nobody flagged
- From click-through prototype to live orders in under two weeks
Building something with SvelteKit or agents?
Tell me what it is and where it's stuck. Email is the quickest way to reach me.