Portfolio
Last updated: September 25th 2026
Iâm Scott, a UK-based product engineer and Svelte consultant with 9 years of commercial web development experience. I build production AI systems, coding-agent infrastructure, MCP tools, SvelteKit products, and open-source developer tooling through OES Technology.
This is selected work rather than every repository Iâve ever pushed. Each project shows the problem, the useful parts, and where you can inspect the source or product.
Selected outcomes
Client work is private, so these are described without names.
Reinsurance underwriting platform (2026). I started from a clickable prototype with no backend and built the real platform, designed to support roughly $20m in business: submission intake, document extraction, rules evaluation, quoting, and audit history. I set up the monorepo and wrote the core web app, worker, rules, database, and domain packages. For the first two months I worked on it alone, writing 520 of the first 523 commits, and in that time it went from first commit to production inside the clientâs private Azure network in about six weeks. The team that joined from late June built on that foundation. After handover, the team raised the platformâs maximum underwriting capacity by more than 130%, an eight-figure increase, in one week by changing rules and configuration, not by rebuilding anything. I also cut the heaviest page response from 50MB to 38KB for a submission with nearly 14,000 locations.
Guardrails that outlast me (2026). Coding agents write plausible code in the wrong place, so on the same platform I made the architecture executable. The boundary checker parses the code and enforces 36 blocking rules, alongside data-ownership, route-data, and lint checks. After handover they still gate the clientâs production deploys, have blocked 8 deploys that broke the rules, and the team has extended them instead of switching them off. The capacity change landed entirely inside the rules, database, and domain packages. The same checker now runs in the CI of a second product.
Context on demand for agents (2026). The projectâs documentation grew to 291 requirements, specs, meeting notes, and client communications. I built a docs search CLI on SQLite FTS5 that extracts decisions, requirements, and risks, so agents pull the right context instead of guessing. Agents called it 1,122 times across 327 sessions, a teammate adopted the approach for another project, and it led to my open-source tool wiki0.
A clean client handover (2026). The client repository is a verified export: they receive 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 through.
Claims platform for deposit-replacement insurance (2026). I took a claims platform from spec to production in about five weeks: filing, adjuster review, tenant responses, video evidence, and valuation. The client was very positive at the demo and plans to train sales staff to sell it. I also maintained the wider live platform, releasing every production change since March and catching failures nobody had flagged, including stopped off-site backups.
Mobile sign-up platform for a telecoms reseller (2026). I turned a click-through prototype into a live production app in under two weeks, and it has taken real customer orders since July. Along the way the test suite went from zero to more than 120 tests, and known dependency vulnerabilities went from nine to zero.
AI customer-service agents (2023â2026). I led 12 developers across two product teams. The agent runtime I built reduced average handle time by 40% through skill-based routing across chat, email, and SMS. I also led an architecture change that resolved 928 build errors and let four applications ship independently.
my-pi
my-pi is my curated Pi distribution: a ready-to-run coding agent CLI with MCP, LSP, skills, recall, redaction, telemetry, team mode, prompt presets, and the bits I want wired in when I work in a terminal.
Key Features:
- Pi-native CLI with interactive TUI, print mode, JSON mode, RPC mode, and a programmable runtime
- Project-aware MCP tools and skills, scoped per repo so noisy or sensitive tools only load where they belong
- LSP-backed diagnostics, hover, definitions, references, and document symbols
- Local SQLite context sidecar, telemetry, evals, recall, and live observability
- Secret safety through output redaction and
nopeekreminders - Team mode for local RPC teammate orchestration with tasks and mailboxes
Tech Stack: TypeScript, Pi, SQLite, MCP, LSP, pnpm
Svortie
An agent orchestration dashboard for defining Claude agents, deploying them into sandboxed environments, and monitoring execution in real-time. Svortie lets you manage the full lifecycle of AI agents from configuration through to live log streaming, all from a single interface.
Key Features:
- âCreate with AIâ: describe what you need in natural language (via Deepgram) and Svortie figures out whether to build an agent, workflow, schedule, or a combination of all three
- Agent management with system prompts, tasks, model selection, and MCP server configuration
- Visual workflow builder for multi-step sequential agent execution
- Real-time execution monitoring with live log streaming
- Sandboxed environments via Daytona SDK for isolated agent runs
- Usage-based billing with credit packs via Polar
- Authentication with email/password and GitHub OAuth
Tech Stack: SvelteKit, Svelte 5, TypeScript, SQLite, Tailwind CSS, shadcn-svelte, Better Auth, Claude Agent SDK, Daytona, Deepgram, Bun
Devhub CRM
A developer-first relationship management platform that solves the core problem developers face with traditional CRMs: they hate manual data entry and generic tools donât fit their unique workflow. Built from the ground up for how developers actually manage relationships - through GitHub, conferences, and open source communities.
Key Features:
- GitHub Quick Connect for seamless contact import from your development network
- Core CRM functionality including contacts, interactions, follow-ups, and notes
- Relationship health scoring to track connection strength
- Public developer profiles with QR codes for easy networking at conferences
Impact:
Designed specifically for conference networking, open source maintainers, developer advocates, and freelancers who need to manage developer relationships without the overhead of traditional CRMs.
Tech Stack: SvelteKit, Svelte 5, TypeScript, SQLite, sqlite-vec, Tailwind CSS, daisyUI, Better Auth
Sveltest
A weekend project that evolved into a comprehensive testing resource for the Svelte ecosystem, now influencing how modern Svelte applications approach testing in real browser environments. Originally created as a companion to my blog post about Migrating from @testing-library/svelte to vitest-browser-svelte, Sveltest has become a community-driven resource built by developers, for developers.
Key Features:
- Comprehensive testing documentation and real-world examples using vitest-browser-svelte
- Pre-configured AI assistant rules for Cursor and Windsurf that enforce testing best practices automatically
- Multi-layer testing approach with minimal mocking - shared validation logic, real FormData/Request objects, and TypeScript contracts
- Lots, and I mean lots of testing scenarios
- Battle-tested patterns from production environments using bleeding-edge Svelte 5 and vitest-browser-svelte
Why it exists:
Svelte testing guidance was spread across documentation, examples, and rapidly changing package APIs. Sveltest puts working browser-mode examples and reusable assistant rules in one public project so teams can inspect and adapt the patterns rather than copy an unverified snippet.
Tech Stack: SvelteKit, TypeScript, Vitest, vitest-browser-svelte, Playwright, TailwindCSS, daisyUI
SkyKit.blue
A weekend project that gained significant traction - SkyKit provides free analytics and statistics for Bluesky users. The application offers post analytics and an inactive account finder with no authentication required, making it super low friction for users.
Key Features:
- Post analytics and engagement metrics
- Following account activity tracking
- Inactive account detection
Tech Stack: SvelteKit, TypeScript, daisyUI
AudioMind
An MP3 to AI Chat Assistant that transforms audio content into interactive, searchable conversations. This project allows users to upload audio files (like podcasts or lectures) and then have natural conversations with the AI about the content.
Key Features:
- Audio file processing and transcription
- Vector-based semantic search
- Multiple AI response styles
- Interactive chat interface
- Real-time progress updates
Tech Stack: SvelteKit 2.x with Svelte 5, TailwindCSS, daisyUI, Anthropic Claude 3, Deepgram, Voyage AI, Turso
Model Context Protocol tools
Iâve built 21 MCP servers and tools since January 2025. On 25 September 2026 they had more than 1,400 GitHub stars, 260 forks, and 11,000 npm downloads a month. The numbers change, so the repository list is the source of truth.
- Omnisearch provides one interface for web search, AI answers, and content extraction across several providers. View the repository.
- SQLite Tools separates read-only, write, schema, and transaction operations for safer SQLite access. View the repository.
- Svelte Docs gives coding agents focused access to current Svelte documentation. View the repository.
- Memory for libSQL provides persistent knowledge and vector search with SQLite and libSQL. View the repository.
- McPick manages MCP servers and skills across coding-agent clients. View the repository.
SvelteKit Embed
A popular package of SvelteKit embed components that makes it easy to add third-party embeds like YouTube, Twitter, and more to your SvelteKit projects.
Key Features:
- Responsive embeds
- Easy component-based implementation
- Support for multiple platforms
Svead
Svelte + Head == Svead! A component that allows you to set head meta information, canonical links, title tags, Twitter and Facebook Open Graph tags, and schema.org data for SvelteKit projects.
Community involvement
- Svelte Ambassador: recognised by the Svelte team;
- Svelte Society London: co-founder and co-organiser;
- Technical writing: more than 240 public engineering articles;
- Speaking: conference talks, workshops, podcasts, and community events about AI engineering, coding agents, Svelte, and SvelteKit.
Technical writing
I publish first-hand engineering guides rather than product summaries. Current highlights include:
- Coding agent harnesses with my-pi
- How I stop LLMs drifting in production codebases
- Building and testing MCP tools locally
- Hardening redaction in my-pi with evals and telemetry
Get in Touch
For product engineering, coding-agent infrastructure, Svelte consulting, workshops, or speaking, use the contact form. My main public profiles are:
- GitHub: @spences10
- Bluesky: @scottspence.dev
- YouTube: Scott Spence Please
Stay Updated
If youâre interested in knowing when I publish new projects or blog posts, you can sign up for my newsletter below.
Sign up for the newsletter
Want to keep up to date with what I'm working on?
Join other developers and sign up for the newsletter.
I care about the protection of your data. Read the Privacy Policy for more info.






