Open to new roles
- Head of Engineering
- Hands-on technical leadership
- AI-first engineering
Remote, or flexible hybrid in the North East · get in touch
$ whoami
Craig Thompson
Head of Engineering · hands-on, AI-first
[email protected] · craigthompson.dev · github.com/Craga89 · linkedin.com/in/craigmichaelthompson · Newcastle upon Tyne, UK
The short CV is at craigthompson.dev/cv.
20+ years’ experience in software development and management, from start-up to acquisition.
I’m a hands-on, full-stack engineer across UI, back end, infrastructure and AI, with extensive engineering leadership experience. I led an engineering department to AI-first development, where most code is now generated by AI, then reviewed by AI and people. Problem solving is what I do best, whether the problem sits in the code, the platform or the team.
I’ve led engineering teams for 12+ years, growing and developing teams of 10+ engineers, both in-house and outsourced. I’ve set strategy, supported the people and steered the platform with them, from monorepos and build engines to AWS, authentication and AI tooling.
I’ve reported directly to CEO and CTO level, turning the roadmap and technical direction into plain-English board updates. I’ve been part of ISO 27001 and GDPR accreditation and the annual audits.
$ ./readme --focus=ai
A suggested reading order for AI-first engineering. Everything else is still here.
- responsibilities/ai.md From AI-assisted to AI-first, in a year
- case-studies/ai/platform.md AI where it earns its keep
- case-studies/this-site.md This site, run like a production system
- case-studies/webmcp.md A site AI agents can use, not just read
- case-studies/testing.md A test suite that kept up with the platform
$ ./readme --focus=frontend
A suggested reading order for front end and accessibility. Everything else is still here.
- case-studies/accessibility.md Accessible by default, checked on every deploy
- case-studies/i18n.md One product in 36 languages, kept in step
- case-studies/refactor.md Keep the toolchain modern, one reviewed change at a time
- case-studies/testing.md A test suite that kept up with the platform
- case-studies/monorepo.md One codebase, kept modern without stopping delivery
$ ./readme --focus=leadership
A suggested reading order for engineering leadership. Everything else is still here.
- responsibilities/leadership.md Led the team, not just the platform
- responsibilities/hiring.md Building the team, from first interview to first deploy
- responsibilities/acquisition.md From startup to acquisition
- responsibilities/budget.md Every pound accounted for
- responsibilities/incidents.md When things break: a plan, not a panic
- case-studies/cost.md Measure first, then cut
$ ./readme --focus=platform
A suggested reading order for platform engineering. Everything else is still here.
- case-studies/monorepo.md One codebase, kept modern without stopping delivery
- case-studies/ci-cd.md Build only what changed, ship every merge
- case-studies/aws.md Nothing public but the CDN
- case-studies/security.md Closing doors, not adding locks
- case-studies/data.md Making the flagship report fast at scale
- responsibilities/incidents.md When things break: a plan, not a panic
$ cat career.log # experience
#Head of Engineering · Hive HR
Led engineering for a B2B SaaS employee engagement platform (surveys, peer-to-peer recognition and confidential feedback) that runs 24/7 for enterprise, mid-market and SME customers, with a cross-functional team of up to 12 engineers.
- 2019–21 · Founding & foundations. Led the consolidation of ~18 repositories into one TypeScript monorepo, with history intact, and set up its , Docker and local dev. Started the React design system and led front-end work on the core product.
- 2021–24 · Platform ownership. Moved CI to self-hosted Buildkite as code, driven by an in-house TypeScript build engine. Helped take the company to , led repo-wide toolchain migrations and hardened AWS access.
- 2024–26 · Architecture, security & AI. Led the move to and an edge behind and with the infrastructure as code in Terraform and CloudFormation, the replacement of our authentication system and an evidence-led push on AWS costs, and helped start an AI agent service.
- 2026 · Acquisition by WorkBuzz. Saw the company through from startup to its acquisition, providing technical leadership along the way and representing engineering right through due diligence.
Technical Lead · Communicator
- Led a Scrum team of junior-to-mid engineers, setting the code quality standards, building the next generation of an email marketing SaaS platform that sends billions of emails, leading front-end development and mentoring two software engineers.
- Architected a React drag-and-drop WYSIWYG Email Designer so non-technical clients could build compatible HTML emails.
- Built a Node.js/Cheerio microservice to automate CSS inlining and HTML optimisation and keep rendering consistent across major email clients.
- Led the team’s move from TFS to Git and TeamCity, then introduced GitLab with runners hosted on our own servers for centralised, faster CI/CD.
- Built Node.js functionality for new products.
Front-end Web Developer · Eutechnyx
Turned visual mock-ups into interfaces and CMS pages for the Auto Club Revolution project with JavaScript and Backbone.js.
Consultant & open-source maintainer · craigsworks
One-off projects for smaller and larger clients worldwide, from single websites to larger builds such as a crop management application, using Node.js, Express and modern SPA frameworks; UI libraries built to WCAG; and maintaining open-source tools like qTip2, from issues and documentation to releases.
$ ls ./responsibilities # roles & responsibilities
#The responsibilities I've held beyond the code, and the impact they had: leading and growing the team, hiring, working with product and customers, incidents, budgets, ways of working and the move to AI-first engineering, seeing a startup through to acquisition, compliance and internal IT. It plays through each file on its own and holds while you hover or read. Jump to any file, or open it on its own page.
$ ls ./case-studies # case studies
#The technical projects I've led with my teams: what we built, what changed, and what it means in practice. Open a file.
// monorepo & developer experience · 2019 → today
One codebase, kept modern without stopping delivery
I’ve taken a platform spread across about 18 separate repositories and led merging them into one TypeScript monorepo with their history intact, along with the tooling around it: CI/CD, Docker and local dev. That one codebase is what makes the repo-wide migrations in refactor.md possible.
2019 · 18 repositories
- 12 services
- React app
- 3 Lambdas
- Terraform
- QA suite
today · one monorepo
100+
packages today, including the shared React design system
# in practice
- Every engineer works in one repo, with one toolchain and one way to build, test and deploy.
- Generated OpenAPI 3.1 docs for 12 services replaced a hand-maintained Postman collection.
// refactors & tooling migrations
Keep the toolchain modern, one reviewed change at a time
Tooling ages fast. Rather than letting it drift, I’ve led a steady run of repo-wide migrations: linting, packages, bundling, types, runtime and the compiler itself. Each one was scripted with a and landed as a single reviewed change, so there was never a long-lived branch or a half-migrated codebase.
Repo-wide migrations I've led9,700+ files moved across them
- lintingTSLint to ESLint455 files
- ciSemaphore to self-hosted BuildkiteCI as code
- packagesyarn to pnpm workspaces619 files
- bundlingWebpack to Vite3 front ends
- typestype aliases to interfaces804-file codemod
- lint + formatESLint + Prettier to oxlint + oxfmt4,366 files
- runtimecompiled JS to TypeScript run natively by Node3,502 files
- compilerTypeScript 5.8 to TypeScript 7~10× faster typechecks
- e2e testingCypress to Playwrightfaster, more stable CI
# in practice
- Upgrades land across the repo in one change, so the codebase doesn’t drift into “old” and “new”.
- Back-end services start straight from TypeScript source, with no build step in between.
- Faster tools where it counts: one quick linter and formatter, and typechecks around 10× faster on TypeScript 7.
// ci/cd & release engineering · 2019 → today
Build only what changed, ship every merge
I built a CI/CD platform from scratch and have kept evolving it as the team and codebase grew. Today it runs on self-hosted Buildkite defined in Terraform, driven by an in-house TypeScript build-and-deploy engine. It models the monorepo’s dependency graph, so each change builds and deploys only the packages it affects.
Every merge to master, end to end
- mergeto master
- package graphwhat's affected?
- build in parallelaffected packages only
- Validate gate
- stagingauto-deploy
- productionlive, no release train
How it evolved: pipelines → a graph-aware release engine
- hosted pipelines
- Buildkite + build engine
- continuous deployment
- merge queue · staging
- one Validate gate
# in practice
- Merged work reaches customers automatically. There has been no release train since 2022.
- A small fix rebuilds a handful of packages, not the whole platform.
- Every PR passes one required quality gate, and
/deployon a PR puts it into a test environment. - Deploys wait for healthy tasks and detect rollbacks, so a bad release stops itself.
// test automation · cypress → playwright
A test suite that kept up with the platform
I led the inception of the first automation suite, built on Cypress, and evolved it in CI over time alongside the Test Automation Engineers. As the suite grew, its performance suffered, so I led the switch to Playwright, which brought huge gains in performance and stability across the CI/CD environment.
How the suite evolved
- first automation suiteCypress
- grown in CIwith Test Automation Engineers
- slower as it grewperformance suffered
- switchedto Playwright
- todayfaster, more stable CI/CD
before · Cypress
The first automation suite. As it grew, its performance in CI suffered.
after · Playwright
Huge gains in performance and stability across CI/CD.
Inception
Led the creation of the first end-to-end automation suite, on Cypress.
Evolving in CI
Grew the suite inside the CI pipeline over time, alongside the Test Automation Engineers.
The switch
When performance suffered as the suite grew, led the move to Playwright.
# in practice
- Automated end-to-end testing has been part of CI since the first suite.
- The suite was built up together with the Test Automation Engineers.
- Moving to Playwright made the suite much faster and more stable in CI/CD.
// accessibility · wcag 2.2 aa · this site
Accessible by default, checked on every deploy
Accessible UI has been part of my work since building UI libraries held to WCAG for clients. This site is held to WCAG 2.2 AA, and an automated suite checks it after every deploy and every week, failing the check if anything regresses.
- audit
- walk it by keyboard
- fix at the source
- gate in CI
What the suite checks
- axe-core · WCAG 2.0–2.2 A/AAEvery automated rule, in every stateBoth themes, desktop and mobile, every tab, open FAQs, search results, ask.sh, the chat, the search dialogs and every file page: 284 page states per run.
- 2.1.1 · 2.4.7Keyboard onlyA Tab-key walk through every focus stop, confirming each one shows a visible focus ring.
- 2.4.11 · new in 2.2Focus never hiddenNothing focused can sit behind the sticky header or the chat bar.
- 2.5.8 · new in 2.2Targets you can hitControls are at least 24px, or spaced so a miss won’t land on a neighbour.
- 1.4.10 · 1.4.12Reflow and spacingNothing scrolls sideways at 320px, and nothing clips with the WCAG text-spacing overrides applied.
- 2.2.2 · 2.3.3Motion under controlReduced-motion users get a still page, and the auto-playing tour can be paused.
- 2.1.4Shortcuts you can switch offThe single-key shortcuts can be turned off, so they never clash with screen readers or speech input.
areabeforeafter
contrastmuted text under 4.5:1Every text colour at 4.6:1 or more on every background it sits on
focussome controls with no ringA visible focus ring on everything that can take focus
obscured focusfocus could hide off screenScroll padding clears the sticky bars; hidden UI leaves the tab order
targets18px links24px minimum
wide tablesunreachable by keyboardFocusable, named scroll regions
# in practice
- Lighthouse alone scored the site 100 for accessibility while these WCAG 2.2 AA failures were present; the full suite found them.
- Fixes went into the shared design tokens and components, so each one holds across the whole site.
- A regression fails the accessibility check, so it can’t quietly slip back in.
- Automated checks cover what tools can see; screen-reader passes cover the rest.
// internationalisation · 2019 → 2026
One product in 36 languages, kept in step
Customers survey their people in their own language, so the interface and the questions have to work in every language they turn on. I’ve kept the platform’s 36 interface languages in step with English, and made translating customers’ own content cache-first and selective.
36
interface languages, including Welsh and Arabic
~116k
blank strings removed, so nothing renders empty
cache first
translations checked inside the VPC before any network call
Interface text, kept in step by the build
- messages in components
- extracted to English
- keys synced to every language
- gaps fall back to English
Customer content, translated at runtime, cache first
- survey question
- Redis cache inside the VPC
- translation service + DynamoDB
- Google Cloud Translation
areabeforeafter
language filesdrifted as keys changedEvery build adds and removes keys in every language to match English
missing textshown as a blankFalls back to the English message
new languageswired in by hand; three droppedGenerated from the language files present
translation callsa network round trip each, often timing outCache inside the VPC first, with one set of timeouts across services
what's translatedevery name, on every read and writeOnly content written to be read in other languages; place names stay as written
language pickereach language in its own scriptEvery language named in the reader’s own language
emojisent to the translator as-isSwapped for keywords and restored, so they come through translation intact
How it evolved: drifting files → in step on every build
- language service (2019)
- locale pipeline rebuilt
- translations in the survey builder
- cache-first translation
- synced on every build
# in practice
- A changed or removed string updates every language in the same change, so no file falls behind.
- Untranslated text shows in English rather than as a gap on screen.
- Adding a language means adding its file; the build generates the rest.
- Customer content is translated only when it should be, and served from a cache next to the services.
- This site follows the same ideas on the visitor’s own device: English fills any gap, translations are cached, and names and file names are never translated. See this-site.md.
// aws architecture · 2021 → 2026
Nothing public but the CDN
I led the platform’s move from containers on EC2 hosts behind a public load balancer to Fargate tasks that talk over , with all customer traffic entering through CloudFront and WAF. Each step shipped without a big-bang cutover.
before · ~2020
- Internet
- Public load balancer · internet-facing
- ECS on EC2 hosts · hosts to patch
- Database via VPC peering
Operator access: SSH bastion
after · 2026
- Internet
- CloudFront + WAF · the only public entry
- Internal load balancer · private
- Fargate · Service Connect · no hosts
- Database via PrivateLink
Operator access: SSO + SSM, no SSH
How it evolved: hosts to patch → no hosts at all
- ECS on EC2 hosts
- capacity autoscaling
- awsvpc networking
- Fargate rollout
- Fargate + Service Connect
# in practice
- No servers to patch: every service runs on Fargate.
- The attack surface is one CDN. WAF moved to blocking only after a 14-day shadow run showed no legitimate traffic would be stopped.
- Each service receives only the events it handles, using subscription filters generated from its code at build time.
- Dev environments share one networking stack and one database cluster, so a new environment is cheap to stand up.
// security & privacy · 2021 → 2026
Closing doors, not adding locks
When a product handles confidential data, security and privacy are part of the product. We focused on removing whole classes of risk rather than layering controls on top.
areabeforeafter
server accessSSH & bastionsSSM Session Manager; port 22 closed everywhere
admin accessIAM admin usersSSO only, via IAM Identity Center and Control Tower
networksVPC peeringPrivateLink and VPC endpoints
public surfaceinternet-facing load balancerOne CDN entry point, with WAF in block mode
dev secretsplaintext .env filesInjected at runtime from a password manager
dependenciesknown CVEs lingeringVulnerable packages swapped out; pentest findings fixed and tracked
incidentsad hocBlameless post-mortems, with runbooks and tooling as the fix
# in practice
- Far fewer standing keys and passwords for anyone to leak.
- No servers reachable from the internet except through the CDN.
- External penetration-test findings were fixed in code and tracked to closure.
// authentication & identity · 2024 → 2026
Sign-in that enterprises trust and users don't notice
I led the replacement of a legacy, multi-tenant sign-in system with a new auth framework, rolled out organisation by organisation behind a per-tenant switch. Before that, we hardened the old one: expiring tokens, central signing and key rotation.
Every sign-in, end to end
- sign-in requestfrom a tenant domain
- resolve organisationfrom the host name
- choose a method
- 2FAenforced per org
- rate-limit checkper identity
- session + signed tokenEdDSA · published keys
rate limiting by identity, not just IP
Separate buckets per user, email, token and IP, per organisation, so one noisy office network can't lock out everyone behind it.
zero-downtime key rotation
Verifiers accept old and new keys during an overlap window, so signing keys rotate without logging anyone out.
How it evolved: legacy service → enterprise-grade sign-in
- legacy auth service
- shared auth library
- expiring, rotated tokens
- new auth framework
# in practice
- Enterprise customers sign in through their own identity provider and can require 2FA for everyone.
- Sign-in links in emails still work after Slack or Outlook previews them.
- Tokens expire and keys rotate on a schedule, with no forced logouts.
- Services can verify tokens themselves against published keys, ready for service-to-service auth.
// data & performance · 2021 → 2026
Making the flagship report fast at scale
Reporting is often the first place a SaaS platform slows down: large aggregations sliced by many dimensions at once. I led the redesign of a core reporting data model and its queries on MongoDB, and we proved the new version matched the old one in production before switching over.
before: work done at read time, on every request
- every response
- unwind all facets
- group everything
- confidentiality in the query
after: work moved to write time, reads stay lean
- facets denormalised on write
- purpose-built collection
- high-cardinality-first index
- analytics-node read
- confidentiality in code, cacheable
10×
higher complexity ceiling for core reports
1–3 s
search stalls removed from people lists under load
~8,000
records no longer loaded to open one dashboard
V1 ⇄ V2
diff mode proved parity in production
How it evolved: read-time work → write-time design
- streamed bulk operations
- reporting rework
- purpose-built collection
- Atlas Search
# in practice
- Large customers can slice results by far more demographics before hitting limits.
- Bulk operations on very large tenants stream records instead of loading them all into memory.
- Every environment, and every laptop, can restore a realistic “gold” dataset in one step.
// cloud cost · 2022 → 2026
Measure first, then cut
Every saving started with data: flow logs, utilisation history or a line in the bill. Each change landed with the numbers that justified it in the pull request.
- measure
- attribute
- change
- verify
Documented monthly AWS savings≈ $1.3k / month from the first three alone
Container Insights: enhanced → standard $620/mo
Duplicate CloudTrail trails removed $450/mo
BI network: 3 AZs and NATs → 1 $270/mo
VPC flow logs off (per account) $250/mo
Also cut, from measured utilisation
- Traced ~1.2 TB/day of CI NAT traffic; replaced managed NAT with endpoints, image caching and fck-nat
- Halved CPU and memory on services idling at 1–3% CPU
- Right-sized Redis after 15 months under 4% CPU
- Non-production environments scale down every night
# in practice
- Spend tracks real usage, and every cut has the evidence behind it on record.
- The NAT investigation became a reusable audit any engineer can rerun.
// ai platform · mastra + bedrock · 2024 → 2026
AI where it earns its keep
I helped introduce AI into the product: an agent service built with , running models through in its own isolated AWS account, with an that runs in CI.
How AI reaches the product
product
features that call AI
in the core platform
isolated AWS account
AI agent service
Mastra agents & workflows
TypeScript, like the rest of the platform
models
Amazon Bedrock
inside the same cloud as everything else
- every change:
- PR to the AI service
- eval suite in CI
- ships only if it passes
# in practice
- AI features are built as agents and workflows in TypeScript, so any engineer on the team can work on them.
- The AI service has its own account, with its own permissions and its own bill, separate from the core platform.
- Changes to the AI service are checked by its eval suite in CI before they reach customers.
// webmcp + mcp · tools for AI agents · 2026
A site AI agents can use, not just read
AI assistants are starting to work inside the browser, on the page a person has open. Rather than leave them to read the screen and guess, I’ve given this site its own tools through WebMCP, a draft web standard. An agent calls a tool and gets structured answers, drawn from the same content visitors read. The same tools also run as an MCP server, so any MCP client can use them without the site open.
In this browser: no WebMCP, so the page skips it.
MCP server https://craga-ask.craga89.workers.dev/mcp
How an agent uses the tools
- AI agent in the browser, or any MCP client
- calls a tool through WebMCP or MCP
- reads the site's own data cv.json, the ask.sh index, CI results
- gets a structured answer or opens it on screen
The tools
get_profileheadline, summary, contact, CV links and skillsget_careerevery role, with highlights, and the degreelist_files·read_filecase studies and responsibilities in full, and can open one on the pagesearch_sitethe best-matching passages, in my own wordsaskan ask.sh answer, with the same limits as the box on the pageget_open_sourcemerged pull requests by quarter, and qTip2get_site_reportthe latest Lighthouse, WCAG and ask.sh resultsshow_sectionscrolls to a section and highlights it (on the page only)
# in practice
- Each tool is defined once, in one shared module, and runs in both places: on the page through WebMCP and on the Worker through MCP. The site’s search lives there too, so ask.sh, the page and the server can’t drift apart.
- The tools read the same files the page loads, so an agent gets the same facts a visitor reads.
- The MCP server is public, read-only and rate-limited. It keeps no state between requests and serves both the current protocol (2026-07-28) and older clients.
- Browsers without WebMCP skip it. In Chrome it runs as an origin trial.
- CI tests both: the page’s tools against a stand-in for the browser API, and the server against both protocol versions and the official MCP client. The ask tool is stubbed in tests, so they never reach the live service.
- For agents that read rather than act, the same content is in llms.txt and cv.json.
// this site · craigthompson.dev · [github.com/Craga89/portfolio](https://github.com/Craga89/portfolio)
This site, run like a production system
It’s a small site, but it’s built and run the way I’d run a product. Everything lives in one public repository, every change goes through automated quality gates, and the AI feature has tests of its own. The results below come live from the latest CI runs.
Latest results
- Lighthouseloads on the live siteperformance, accessibility and best practices, mobile and desktopfrom the latest CI run · view run
- WCAG 2.2 AA suiteloads on the live siteaxe-core in every page state, plus keyboard, focus, reflow and motion checksfrom the latest CI run · view run
- ask.sh answer testsloads on the live siteright facts, nothing invented, no prompt injectionfrom the latest CI run · see the tests
How a visit works
- the page Browser GitHub Pages static, prerendered, works offline
- ask.sh Browser Cloudflare Worker rate-limited per visitor, and capped within the free daily AI allowance Workers AI three models, each a fallback for the one before; answers only from this site
- counts Browser Cloudflare Worker D1 anonymous question log and visit counts
- agents AI agent in the browser WebMCP or MCP the same tools, on the page or on the Worker: profile, career, files, search and ask.sh
- translation Browser its built-in Translator API, when the visitor wants another language on the device 38 languages, right-to-left included; nothing is sent anywhere
What every change goes through
- qualitylint, format and type checks, and unit tests with a coverage threshold
- buildprerender the page, self-host fonts, resize images, print cv.pdf, write the contact card and the ask.sh index
- testend-to-end tests on a desktop and a phone, the WCAG 2.2 AA suite and a screenshot comparison against approved baselines
- deployGitHub Pages, only when every gate passes, stamped with the commit; then Lighthouse thresholds
- workerdeploy, a smoke test, then the ask.sh answer tests
- weeklyevery check again, plus a dead-link check, visit and question stats, and the open-source counts
# in practice
- The same gates I’d put on a product: a change that hurts accessibility, performance or the visuals fails its check.
- The AI feature is tested like any other code, for the right facts, invented ones and prompt injection.
- Anonymous counts instead of tracking: no cookies, no IP addresses, nothing that identifies a visitor.
- It runs on the free tiers of GitHub Pages and Cloudflare.
$ ls ./stack # skills
#87 tools across 12 areas · type to search
languages/ 7 TypeScript, JavaScript, HTML, CSS, Node.js, Bash, HCL
- TypeScript
- JavaScript
- HTML
- CSS
- Node.js
- Bash
- HCL
backend/ 7 Express, GraphQL (Apollo), Koa, REST / OpenAPI, Zod, RxJS, Microservices
- Express
- GraphQL (Apollo)
- Koa
- REST / OpenAPI
- Zod
- RxJS
- Microservices
data/ 5 MongoDB Atlas, Atlas Search, Elasticsearch, Redis, DynamoDB
- MongoDB Atlas
- Atlas Search
- Elasticsearch
- Redis
- DynamoDB
frontend/ 8 React, Apollo Client, Emotion, Vite, Storybook, Playwright, Cypress, WCAG
- React
- Apollo Client
- Emotion
- Vite
- Storybook
- Playwright
- Cypress
- WCAG
compute/ 8 ECS, Fargate, Service Connect, EC2, Auto Scaling, Lambda, Batch, ECR
- ECS
- Fargate
- Service Connect
- EC2
- Auto Scaling
- Lambda
- Batch
- ECR
networking/ 9 VPC, CloudFront, WAF, Route 53, Elastic Load Balancing, API Gateway, PrivateLink, VPC endpoints, NAT / fck-nat
- VPC
- CloudFront
- WAF
- Route 53
- Elastic Load Balancing
- API Gateway
- PrivateLink
- VPC endpoints
- NAT / fck-nat
messaging/ 2 SNS, SQS
- SNS
- SQS
access/ 7 IAM, IAM Identity Center, Control Tower, Organizations, Systems Manager (Session Manager), CloudTrail, VPC Flow Logs
- IAM
- IAM Identity Center
- Control Tower
- Organizations
- Systems Manager (Session Manager)
- CloudTrail
- VPC Flow Logs
delivery/ 8 Git, Terraform, CloudFormation, Packer, Docker / BuildKit, Buildkite, GitHub Actions, pnpm
- Git
- Terraform
- CloudFormation
- Packer
- Docker / BuildKit
- Buildkite
- GitHub Actions
- pnpm
it/ 10 Google Workspace, MDM, Mosyle, Jamf, Apple Business Manager (ABM), macOS fleet, ThreatDown, IT service desk, Jira Service Desk, Status page
- Google Workspace
- MDM
- Mosyle
- Jamf
- Apple Business Manager (ABM)
- macOS fleet
- ThreatDown
- IT service desk
- Jira Service Desk
- Status page
observability/ 7 Sentry, OpenTelemetry, Fluent Bit, Grafana, CloudWatch, Container Insights, AWS X-Ray
- Sentry
- OpenTelemetry
- Fluent Bit
- Grafana
- CloudWatch
- Container Insights
- AWS X-Ray
ai/ 9 Claude Code, Anthropic API, Amazon Bedrock, Amazon Q, Mastra, promptfoo, Pi, Superset, GitHub Copilot
- Claude Code
- Anthropic API
- Amazon Bedrock
- Amazon Q
- Mastra
- promptfoo
- Pi
- Superset
- GitHub Copilot
grep: no matches. Ask me anyway: [email protected]
$ ls ./open-source # open source
# Q1Q2Q3Q4 2012 Q1 · none 2012 Q2 · none 2012 Q3 · 1 merged: andris9/highlight 2012 Q4 · none
'12 2013 Q1 · 1 merged: cdnjs/cdnjs 2013 Q2 · none 2013 Q3 · 1 merged: cdnjs/cdnjs 2013 Q4 · 4 merged: jsdelivr/jsdelivr, cdnjs/cdnjs, Mottie/tablesorter
'13 2014 Q1 · 4 merged: etodanik/veinjs, itsjavi/bootstrap-colorpicker 2014 Q2 · 7 merged: BlueSpire/Durandal, dobtco/jquery-resizable-columns, giggio/npm-nuget 2014 Q3 · 1 merged: jsdelivr/jsdelivr 2014 Q4 · none
'14 2015 Q1 · 3 merged: codesuki/react-d3-components, aurelia/skeleton-navigation, rniemeyer/knockout-delegatedEvents 2015 Q2 · none 2015 Q3 · 4 merged: mbest/knockout.punches, jakubburkiewicz/uncss-brunch, BlueSpire/Durandal-Bower 2015 Q4 · 2 merged: cdnjs/cdnjs
'15 2016 Q1 · 2 merged: preactjs/preact, jakubburkiewicz/uncss-brunch 2016 Q2 · none 2016 Q3 · 2 merged: unindented/stats-webpack-plugin, shipshapecode/tether 2016 Q4 · 2 merged: electrode-io/electrode-webpack-reporter
'16 2017 Q1 · 7 merged: communicator/dockerized-magento, camwest/react-slot-fill, robinweser/fela 2017 Q2 · 3 merged: ianstormtaylor/slate 2017 Q3 · 3 merged: KoryNunn/scroll-into-view, faceyspacey/babel-plugin-dual-import, lerna/lerna 2017 Q4 · 3 merged: camwest/react-slot-fill, tomkp/react-split-pane
'17 2018 Q1 · 2 merged: communicator/magento-plugin 2018 Q2 · 1 merged: gregberge/recompact 2018 Q3 · 1 merged: Popmotion/popmotion 2018 Q4 · none
'18 2019 Q1 · none 2019 Q2 · none 2019 Q3 · 1 merged: slefevre/gitin2it 2019 Q4 · none
'19 2020 Q1 · none 2020 Q2 · none 2020 Q3 · none 2020 Q4 · none
'20 2021 Q1 · none 2021 Q2 · none 2021 Q3 · none 2021 Q4 · none
'21 2022 Q1 · 1 merged: swc-project/swc-node 2022 Q2 · 1 merged: buildkite/test-collector-javascript 2022 Q3 · none 2022 Q4 · none
'22 2023 Q1 · none 2023 Q2 · none 2023 Q3 · none 2023 Q4 · none
'23 2024 Q1 · none 2024 Q2 · none 2024 Q3 · none 2024 Q4 · none
'24 2025 Q1 · none 2025 Q2 · none 2025 Q3 · none 2025 Q4 · none
'25 2026 Q1 · none 2026 Q2 · 2 merged: shuding/nextra, better-auth/better-auth 2026 Q3 · none 2026 Q4 · none
'26
as a table
| year | merged | projects |
|---|---|---|
| 2012 | 1 | andris9/highlight |
| 2013 | 6 | cdnjs/cdnjs, jsdelivr/jsdelivr, Mottie/tablesorter |
| 2014 | 12 | etodanik/veinjs, itsjavi/bootstrap-colorpicker, BlueSpire/Durandal, dobtco/jquery-resizable-columns, giggio/npm-nuget, jsdelivr/jsdelivr |
| 2015 | 9 | codesuki/react-d3-components, aurelia/skeleton-navigation, rniemeyer/knockout-delegatedEvents, mbest/knockout.punches, jakubburkiewicz/uncss-brunch, BlueSpire/Durandal-Bower, cdnjs/cdnjs |
| 2016 | 6 | preactjs/preact, jakubburkiewicz/uncss-brunch, unindented/stats-webpack-plugin, shipshapecode/tether, electrode-io/electrode-webpack-reporter |
| 2017 | 16 | communicator/dockerized-magento, camwest/react-slot-fill, robinweser/fela, ianstormtaylor/slate, KoryNunn/scroll-into-view, faceyspacey/babel-plugin-dual-import, lerna/lerna, tomkp/react-split-pane |
| 2018 | 4 | communicator/magento-plugin, gregberge/recompact, Popmotion/popmotion |
| 2019 | 1 | slefevre/gitin2it |
| 2022 | 2 | swc-project/swc-node, buildkite/test-collector-javascript |
| 2026 | 2 | shuding/nextra, better-auth/better-auth |
open-source/qtip2
qTip2 — pretty powerful tooltips
Widely adopted jQuery plugin and a go-to tooltip library in its day: ~2k stars, 458 forks and 1,384 commits, built with extensive help from community contributors. Featured in Smashing jQuery (2011).
open-source/micromouse
Micromouse simulator
Interactive maze-solving simulation using the classic flood-fill algorithm, with the Arduino code to run it.
communicator/email-designer
WYSIWYG Email Designer
React drag-and-drop builder with a Node.js CSS-inlining service behind it, for a platform sending billions of emails.
59 merged pull requests across 35 other projects since 2012, usually fixing something I hit in production. A selection:
- Git worktree support for the docs framework shuding/nextra PR #4995
- Fixed SAML metadata so an organisation's default SSO provider works better-auth/better-auth PR #9398
- Path prefixes so Test Analytics works in monorepos buildkite/test-collector-javascript PR #30
- Made the require hook revertible swc-project/swc-node PR #647
- IE11 compatibility fix for the animation library Popmotion/popmotion PR #506
- Custom pane class names and CSS order support tomkp/react-split-pane PR #226
- Improved canary releases for the monorepo tool lerna/lerna PR #904
- Selection and node-lookup utilities for the rich-text editor ianstormtaylor/slate PR #728
- Clearer display names for themed React components robinweser/fela PR #193
- Positioning for elements inside iframes shipshapecode/tether PR #179
- Fixed handling of multiple camelCase CSS properties preactjs/preact PR #66
- Observable app titles and an optional view plugin for the SPA framework BlueSpire/Durandal PR #486
- Published qTip2 releases to jsDelivr jsdelivr/jsdelivr PR #203
- Published qTip2 releases to cdnjs cdnjs/cdnjs PR #1071
# used by My jQuery plugins were used on sites by Mozilla, Salesforce, Lloyds TSB, ASOS, Syngenta, Urban Dictionary, Northumbria University, Skrill, models.com and ThinkTag.
$ ls -t ./blog # writing
#26 posts from my old blog on craigsworks.com, 2012–2015, kept as they were written. Some of the tools have since moved on.
- Generating Roman Numerals generating-roman-numerals.md JavaScript · 1 min
- Tether.js - Transition small movements tether-js-transition-small-changes.md 1 min
- Durandal and Webpack: Introduction durandal-and-webpack-introduction.md Knockout.js · Durandal · jQuery · 2 min
- Durandal and Webpack: Part 1 - Module format durandal-and-webpack-part-1-module-format.md Knockout.js · Durandal · jQuery · 2 min
- Durandal and Webpack: Part 2 - Composition durandal-and-webpack-part-2-composition.md 4 min
- GitHub Pages + Webpack Chunks = output.publicPath github-pages-and-webpack-2.md 1 min
- gulp-watch-less - Watch LESS @imports gulp-watch-less.md Plugins · JavaScript · Gulp · 1 min
- LESS - Generate a deep list of all a file's @imports less-file-import-list.md JavaScript · LESS · 1 min
- Bootstrap 3 Buttons and the enable binding knockout-bootstrap-3-buttons-and-the-enable-binding.md JavaScript · Bootstrap · Knockout.js · 1 min
- Durandal - Type cast queryString parameters durandal-type-cast-querystring-parameters.md 1 min
- Knockout value binding + HTML5 number input knockout-value-binding-input-number.md JavaScript · Knockout.js · HTML5 · 1 min
- ui.draggable.exclusive - Exclusive draggables jquery-ui-draggable-exclusive-draggables.md Plugins · JavaScript · jQuery · 3 min
- CKEDITOR key event - modifier keys ckeditor-key-modifiers.md Plugins · JavaScript · CKEditor · 1 min
- Bootstrap 3 - Disabled dropdown links bootstrap-3-disabled-dropdown-links.md JavaScript · CSS · Bootstrap · 1 min
- jQuery UI Draggable - Droppables in an iFrame jquery-ui-draggable-droppables-in-an-iframe.md Plugins · JavaScript · jQuery · 1 min
- qTip2 - Now on jsDelivr qtip2-now-on-jsdelivr.md Plugins · JavaScript · qTip2 · 1 min
- 60fps scrolling using pointer-events: none 60fps-scrolling-using-pointer-events-none.md 1 min
- qTip2 - v2.2.0 released! qtip2-v2-2-0-released.md javascript · release · plugins · 1 min
- Scratchpad.io – Real-time HTML/CSS rendering scratchpad-io.md 1 min
- qTip2 - Pre-release changes (AKA What happened to my qTip styles!?) qtip2-pre-release-changes-aka-what-happened-to-my-qtip-styles.md plugins · javascript · 1 min
- notyfy = noty + CSS themes noty-css-themes.md Plugins · JavaScript · notyfy · 1 min
- qTip2 – Tip plugin’s dimension options updated qtip2-tips-plugins-dimension-options-changed.md Plugins · JavaScript · 1 min
- var, commas and equals var-commas-and-equals.md javascript · 1 min
- JavaScript iOS version detection javascript-ios-version-detection.md javascript · ios · snippet · 1 min
- qTip2 - Styles update qtip2-styles-update.md plugins · javascript · 1 min
- Efficient JavaScript Delta-encoding – Merging JSON objects efficient-javascript-delta-encoding-merging-json-objects.md JavaScript · Old · Delta Encoding · 2 min
$ cat recommendations.txt # recommendations
#From LinkedIn, quoted as written. See them on LinkedIn →
I worked with Craig over a number of years whilst we were both in the development team at Communicator. Craig's passion towards getting the best solution in place is unmatched. He cares about both the smaller details and bigger picture. I enjoyed working with Craig on a number of projects, we developed a great rapport and I would happily highlight that Craig went from colleague to friend whilst we worked on a number of fantastic features that were added to the Communicator platform.
Craig keeps himself ahead of the game, he understands all of the bleeding edge technology, when it would be worthy of considering as a developmental option and when it's still too much in its infancy to risk adopting. I'd recommend him to anyone and believe his front end skillset is of a substantially high quality.
Even after a number of years separated professionally, I still call Craig a friend and wouldn't think twice about accepting the invitation of a catch up over a cold beverage. I'd also jump at the chance to collaborate again in the future and I really hope our paths will cross again!
Lloyd Rodrigues Powell
Head of Software Development at Northumbria Healthcare NHS Foundation Trust
May 2021 · worked with Craig on the same team
Craig is a very able developer, who is capable of putting together high-quality front-end solutions at lightning speed. Always friendly and willing to lend his expertise JavaScript, jQuery, CSS and related web technologies, he is indispensable part of the Auto Club Revolution web team.
Joe Forster
Senior AI Programmer at Ubisoft Reflections
October 2012 · worked with Craig on the same team
At the time of writing I am working with Craig on an ongoing very large and complex web-development project. Craig is the go-to person for all of the frontend jQuery and CSS related issues. He is also increasingly getting involved with the backend-development. An all-round excellent web-developer.
Peter Davies
Software Engineering Leader | Product Business | Technology
October 2012 · worked with Craig at different companies
$ cat recognition.txt
- qTip featured in Smashing jQuery (2011)
- Famous Software Award for qTip2
$ cat education.txt # education
MEng, Electronic & Computer Engineering
Newcastle University · 2008 — 2013
$ cat faq.md # faq
#What roles are you open to?
Hands-on technical leadership roles, such as Head of Engineering, where I stay close to the code and to AI-first delivery while setting direction for the team. I’m most at home in start-ups and scale-ups.
Remote, hybrid or office?
Primarily remote. I’d consider hybrid for a North East–based company with genuinely flexible attendance, but not a fixed “X days a week” commitment.
Where are you based?
Newcastle upon Tyne, UK.
What's your notice period?
Three months.
How big a team have you led?
Up to 12 engineers as Head of Engineering, and before that a Scrum team of junior-to-mid engineers as a Technical Lead.
Have you hired and grown engineers?
Yes. I’ve hired 10+ engineers, through an interview process I designed with the CTO and Deputy Head. Many of the people I’ve led have been promoted, including one from Software Engineer to my Deputy Head of Engineering. See hiring.md and people.md.
Have you worked with customers and the board?
Yes. I’ve joined sales, customer and due-diligence calls as the technical expert, been the primary escalation point for major customer issues alongside my Deputy, and reported to the CEO and CTO with board material in plain terms. See customers.md and product.md.
How do you use AI in engineering?
Over the past year I’ve led an engineering department from introducing AI-assisted development to a point where the vast majority of code is AI-generated, then reviewed by AI alongside human reviewers against style and coding guidelines iterated on in the repo. See ai.md.
Have you owned a budget?
Yes. I’ve owned the substantial monthly AWS budget for a SaaS platform, and managed the platform’s technical suppliers and the wider IT suppliers, negotiating renewals and contracts. See budget.md.
Have you led incidents and on-call?
Yes. I set up the on-call rota and status page, led on incidents from single-client issues to full outages, and helped produce the business’s disaster recovery process. See incidents.md.
Do you cover internal IT too?
Yes. I’ve been the point person for IT: the hardware estate, Google Workspace, the MDM that provisions and secures the Mac estate, and an IT service desk I set up for everyone. See it.md.
Have you worked with outsourced or offshore teams?
Yes. As Head of Engineering I worked directly with an outsourcing firm in India, onboarding 5+ Software Engineers and Automation Test Engineers to bolster the in-house team. Many of them stayed for 2+ years.
Have you been through an acquisition?
Yes. I’ve led engineering from startup through to acquisition, and represented engineering right through due diligence. See acquisition.md.
Are you still hands-on?
Yes. Alongside setting strategy and leading the team, I’ve remained a skilled, hands-on engineer and upskilled in AI-assisted development along the way. Several of the upstream pull requests on this page are recent.
What industries have you worked in?
Mostly B2B SaaS for enterprise, mid-market and SME customers, including Hive HR’s employee engagement platform and an email marketing platform sending billions of emails, plus consulting for clients worldwide.
Do you have compliance experience?
Yes: GDPR and ISO 27001 accreditation, the annual audits, and arranging the yearly penetration test. See compliance.md.
Which clouds do you know?
Primarily AWS, with some GCP and more limited exposure to Azure.
Can I get a copy of your CV?
Yes. There’s a plain version at craigthompson.dev/cv, also as a PDF and as JSON Resume, or email me and I’ll send it.
Do you track visitors?
No cookies, and nothing that identifies you. The site counts visits (the page, the country, the site you came from and which sections you saw), using its own counter and Cloudflare Web Analytics, and keeps ask.sh questions anonymously, so I can see what’s useful and fill any gaps. Both respect Do Not Track and Global Privacy Control.
$ ./ask.sh # ask about my experience
#Ask anything about my experience, roles or skills. An AI model answers using only what's on this site and links to where it found it. Questions are kept anonymously, with nothing that identifies you, so I can improve the site.
Pick a question below or type your own. Type help for commands.
$ ./ask.sh --test # how answers are checked
After every change to the Worker, and every week, a script asks ask.sh these questions and checks each answer: the right facts, nothing invented, and no response to instructions slipped into a question. One wrong answer fails the run. The tests are in the repo.
The latest results load from CI on the live site.
$ cat get-in-touch.md # contact
#Let's talk
I'm open to hands-on technical leadership roles. If you're hiring, or want to talk about anything on this site, email me.
- Looking for
- Head of Engineering, hands-on technical leadership, AI-first engineering
- Working
- Remote, or flexible hybrid in the North East
- Based in
- Newcastle upon Tyne, UK
- Notice period
- Three months