Scott Spence

Summary

Product engineer building client-facing platforms, internal tools, and AI-assisted engineering workflows across SvelteKit, TypeScript, cloud infrastructure, and LLM tooling. Architected a live reinsurance underwriting platform built to support roughly $20m in business, with the codebase designed from the outset for safe, high-standard AI agent collaboration.

Highlights

Took a live reinsurance underwriting platform from a clickable prototype with no backend to production in the client’s private Azure network in about six weeks, working alone for the first two months (520 of the first 523 commits). Set up the monorepo and wrote the core web app, worker, rules, database, and domain packages that the team joining later built on. After handover, the team raised the platform’s maximum underwriting capacity by more than 130%, an eight-figure increase, in one week through rules and configuration changes, not a rebuild

Owned and maintained a live UK deposit-replacement insurance platform, releasing every production change since March and catching failures nobody had flagged, including off-site backups that had stopped and PDF generation breaking in production. Built its claims platform from spec to production in about five weeks. The client was very positive at the demo and plans to train sales staff to sell it

Built deterministic guardrails so coding agents could not erode the architecture: 36 blocking module-boundary rules parsed from the code, plus data-ownership, route-data, and lint checks. After handover they still gate the client’s production deploys, have blocked 8 deploys that broke the rules, and the team has extended them rather than switching them off. The same checker was adopted in a second product and wired into its CI

Built a docs search CLI (SQLite FTS5, fact extraction, zero dependencies) so agents pull the right requirements, schema notes, and client decisions on demand instead of re-reading a growing docs folder. Agents called it 1,122 times across 327 sessions as the corpus grew to 291 documents and 4,307 extracted facts. A teammate adopted the approach for another project, and it led to my open-source tool wiki0

Designed the client release as a verified export: the client receives exactly the application they bought, while the docs corpus, agent skills, and delivery tooling remain agency IP. The export fails if internal paths or references leak into the client repository

Implemented cloud deployment and operational changes across AWS and Azure, including ECS/container app services, RDS/Postgres-backed applications, SES/email infrastructure, storage, secrets/environment configuration, Terraform PRs, DNS, and production debugging

Used browser automation, MCP tools, recall databases, documentation search, and agentic coding workflows to test production journeys, diagnose failures, coordinate parallel work, and preserve review control while moving quickly

Summary

Engineering team lead for a SvelteKit monorepo ecosystem powering AI customer service agents. Led 12 developers across Customer Portal (8) and Smart Agent (4) teams. Primary architect for infrastructure, security, and AI integration decisions across the platform.

Highlights

Built claude-sdk-runtime, an AI customer service agent deployed on AWS ECS for enterprise clients. Reduced average handle time (AHT) by 40% through intelligent skill-based routing across chat, email, and SMS channels

Pioneered AI agent orchestration workflows with Claude Code. Built reusable engineering skills and coordinated multi-agent teams for parallel development across security, infrastructure, and architecture workstreams

Led a major UI architecture decoupling, migrating the primary application from a shared monorepo UI package to shadcn-svelte. Resolved 928 build errors and unblocked 4 dependent applications to ship independently

Migrated the agent builder to Svelte 5, establishing modern reactivity patterns that Claude Code adheres to when generating new code. This reduced bad practices creeping back into the codebase and freed up the team to move faster

Proactively evaluated Vite 8 / Rolldown / OXC toolchain ahead of release, benchmarking build performance and plugin compatibility to prepare the team for migration

Built reusable security middleware for ownership validation across all API routes, and automated GDPR-compliant trace cleanup for AI observability data

Summary

Independent consultancy providing contract engineering leadership, developer relations, and Svelte expertise to technology companies.

Recognised by the Svelte core team as a Svelte Ambassador for sustained community contributions and helping grow a welcoming ecosystem. Co-founder and organiser of Svelte Society London, running monthly community events for 5 years. International conference speaker at Connect Tech (Atlanta), CityJS, Modern Frontends, NXT Nordics (Oslo), and Jamstack Conf.

Technical content creator at scottspence.com with 1.4M+ page views, 791K+ unique visitors across 244 posts, and 22K+ monthly readers. Writing about SvelteKit, TypeScript, AI tooling, and developer experience.

Author of 20+ MCP (Model Context Protocol) tools with 1,200+ combined GitHub stars, covering search, memory, workflow automation, and documentation access.