Engineering Lead | AI Product Engineer | Svelte Ambassador
AI engineering leader and full-stack developer with 6 years of TypeScript experience and 8 years in commercial development teams.
I build AI-powered products, from customer service agent runtimes that
reduced average handle time by 40%, to multi-agent orchestration
systems and MCP tooling. I ship fast, lead teams, and care deeply
about developer experience.
Currently product engineer at Cloud Lobsters, building AI-adjacent
product systems, client platforms, automation, and internal agent
workflows across SvelteKit, TypeScript, cloud infrastructure, and LLM
engineering. Previously engineering team lead at XtendOps, leading 12
developers across two product teams in a SvelteKit monorepo powering
AI customer service agents.
Co-founder of Svelte Society London (monthly events for 5 years). International conference speaker.
Technical blogger at scottspence.com with 1.4M+ page views and 791K+
unique visitors.
Previous roles include Developer Relations at Storyblok, Developer
Advocate at GraphCMS (now Hygraph), and Web Developer at Karmarama and
Zaizi. See early web development roles for
details.
Prior career of 12 years as a senior-level Analyst Developer, building Microsoft
Office automation for large financial organisations.
Location: UK (Swanley, Kent)
Product Engineer
Cloud Lobsters Mar 2026 - Present (3mos) Summary
Product engineer building client-facing platforms, internal tools, and
AI-assisted engineering workflows across SvelteKit, TypeScript, cloud
infrastructure, and LLM tooling. Built a greenfield insurance workflow
platform intended to process roughly $20m in business when fully live,
with the codebase designed from the outset for safe, high-standard AI
agent collaboration.
Highlights
Built a multi-tenant insurance workflow platform from scratch,
covering submission intake, document processing, extraction/evaluation
flows, quote lifecycle support, operational tooling, audit history,
and controlled deployment into client infrastructure
Designed the repo for agent-assisted team delivery: canonical agent
onboarding, local project skills, branch workflow rules,
module-boundary guidance, service-layer patterns, code-style rules,
MVP readiness gates, and explicit “read docs before changing code”
entrypoints
Created a local documentation package that indexes project docs into a
SQLite/FTS search layer, giving agents fast, cited access to business
requirements, schema notes, communications, workflow maps, meeting
notes, and implementation context before touching code
Added guardrail tooling to keep standards high, including import and
module-boundary checks, route data audits, runtime configuration
checks, release verification/export tooling, required Svelte/type
checks, and guidance for keeping routes thin and business logic inside
services/packages
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
Engineering Team Lead
XtendOps Sep 2023 - Feb 2026 (2yrs 5mos) 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
AI & Svelte Consultant
OES Technology Ltd Apr 2021 - Present (5yrs 2mos) 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.
Projects
Pi Coding Agent:
terminal coding-agent harness with session recall, MCP/tooling
integration, skills, observability, theming, and multi-agent workflows.
Used heavily for real product engineering work and for developing
safer AI-assisted delivery patterns.
Svortie: AI agent orchestration platform
with sandboxed execution via Daytona and Claude Agent SDK. Run
autonomous coding agents safely in isolated environments.
Sveltest: Svelte testing documentation
built to generate AI coding assistant rules for CLAUDE.md, Cursor, and
Windsurf. Boosted team productivity by giving AI agents accurate
Svelte testing patterns.
MCP Ecosystem:
20+ Model Context Protocol tools with 1,200+ combined GitHub stars.
Includes mcp-omnisearch (multi-provider search), mcp-memory-libsql
(persistent memory), mcp-n8n-builder (workflow automation), and
mcp-svelte-docs.
Svead: SvelteKit SEO
metadata component, used by Ashley Furniture and other production
sites.
Interests
AI tinkering and building side projects. Currently exploring Claude
Code multi-agent workflows, sandboxed evals, and MCP tooling.
Writing how-to’s and guides on scottspence.com.
Mechanical keyboards, a never ending rabbit hole!
Watching YouTube videos on mechanical engineering and restorations.
Massive Drum and Bass fan!