Skip to content
Artemis Axen

Artemis Axen — Freelance Shopify & Next.js Developer

Full-stack developer who ships products end to end — and knows what they cost to run: database IO, cache writes, bot traffic and search visibility.

living and working in the Netherlands
18 yrs

living and working in the Netherlands

first commit as a professional developer
2014

first commit as a professional developer

products in production in 2026
4

products in production in 2026

localised pages served on Vercel
30k+

localised pages served on Vercel

Selected Work

01 / Full-stack platform

LiveFeb 2026 – now · 982 commits

LiveCity.cam

A multilingual directory of live webcams worldwide — beaches, cities, airports, volcanoes, aurora. Started as a hobby, grew into a production platform with real search traffic.

livecity.cam
LiveCity.cam homepage with live camera discovery
LiveCity.cam mobile homepage
LiveCity.cam camera page with local time and weather
live cameras
1,500+

live cameras

locales incl. RTL
13

locales incl. RTL

localised pages
30k+

localised pages

Google clicks, 28-day period
+78%

Google clicks, 28-day period

The problem

Serve ~30,000 localised pages from a small Postgres instance and a Vercel budget — while streams go offline every few days and crawlers hammer the site.

Stack

  • Next.js 16
  • React 19
  • TypeScript
  • Supabase / Postgres
  • Vercel ISR
  • YouTube Data API
  • Python batch pipeline
  • Leaflet + CARTO maps

Engineering decisions

  1. 01ISR strategy: English prerendered at build, other locales on demand, 30-day revalidation and tag-based invalidation after content deploys.
  2. 02Catalogue cached in 200-row chunks after the list silently exceeded the 2 MB data-cache limit and production showed zero cameras.
  3. 03Homepage payload cut from 219 KB to 61 KB gzipped (3.6×) by moving the full camera list to a CDN-cached API; TTFB ~1 s → 0.3–0.5 s.
  4. 04Disabled unthrottled Link prefetch and extended revalidation — daily Vercel usage down ~35%.
  5. 05Blocked a scraper responsible for 71% of requests at the Vercel firewall after it exhausted the database IO budget.
  6. 06Python batch pipeline: YouTube Data API checks for duplicates, live status and embeddability, preflight before insert, translation-completeness verification across 13 locales.
  7. 07Found pages that had revalidate but no generateStaticParams — /country, /live-webcams, /blog, blog posts and multi-screen ran a function on every view instead of being cached. Fixed; production now returns x-vercel-cache: HIT.
  8. 08Automatic failover: when the YouTube player reports a stream as gone, the tile switches to a backup camera in the same time zone, so a dead stream needs no deploy. Tested locally against a real dead stream.
  9. 09Search driven by data, not guesses: hand-written titles for low-CTR pages chosen from real Search Console queries (one city page went from 8.3% to 23.5% CTR), and noindex on ~15k thin single-camera area pages to focus crawl budget on strong ones.
  10. 10Stream health check across 1,620 YouTube streams: 86 dead streams swapped to their new live IDs (37 cameras back online after weeks offline), 25 with no live source taken offline.
  11. 11Shipped in one week: reusable event hubs (New Year's Eve with 37 cities in midnight order, Christmas markets), a random-camera page with swipe and a full-screen Reels mode on phones, and hand-written city guides in 13 languages.
  12. 12Postmortem — a cache stampede we caused: purging the catalogue tag twice in one day made every page refetch 1,500+ heavy rows at once. ~169k requests stalled the database, and functions waiting up to 300 s billed ~6× a normal day in provisioned memory. Fixes shipped: a lean catalogue rebuilt with one fetch (heavy fields load per page), a ~10 s database timeout to fail fast, at most one purge a day, batched, tested deploys — and functions moved from Washington to Frankfurt, next to the database, so queries no longer cross the Atlantic. The fix passed its first real purge: sampled pages answered in 0.15–0.7 s with no errors in the logs.

Outcome — Grew from ~190 to 1,500+ cameras in two months through a semi-automated pipeline. Approved for Google AdSense; organic traffic growing month over month.

02 / Content platform

LiveMay 2026 – now

TravelHub.cam

A travel video platform built around my own 4K walking tours, webcam recordings and ferry trips — each video becomes an editorial journey page with route map, chapters and a city essay.

travelhub.cam
TravelHub.cam homepage
TravelHub.cam mobile homepage
TravelHub.cam walking-tour journey page
videos imported
237

videos imported

locales
12

locales

translation rows
693

translation rows

commits since May
285

commits since May

The problem

A large multilingual catalogue on a young domain: Search Console showed ~1,200 thin, untranslated pages diluting the whole site.

Stack

  • Next.js 16
  • React 19
  • TypeScript
  • Tailwind 4
  • Supabase Auth + RLS
  • proxy.ts
  • YouTube OAuth
  • GA4 Consent Mode

Engineering decisions

  1. 01Locale handling in proxy.ts: detection, 404 for unknown locales, 308 to English for untranslated pages, x-locale header for the correct <html lang>.
  2. 02A content-only locale pattern so a single translated article doesn't spawn hundreds of empty pages.
  3. 03Sitemap and indexing limited to editorially processed pages; everything else noindex.
  4. 04Review-then-apply pipeline: AI drafts copy, a human approves, scripts geocode and write — nothing auto-publishes.
  5. 05OAuth write-back that adds deep links into YouTube descriptions, plus a JSON feed consumed by LiveCity.
  6. 06Google sign-in, comments with moderation and reactions on Supabase with row-level security.

Outcome — Clean index, a working editorial pipeline, and two sister sites that feed each other traffic and content.

03 / Content platform · MVP

Live MVPJul 2026 – now

FamilyFound

A knowledge platform for people building families outside the traditional route — co-parenting, known donors, IVF, surrogacy — across the Netherlands, UK, Canada, Australia and the US.

familyfound.io
FamilyFound homepage
FamilyFound mobile homepage
FamilyFound resources library on mobile
guides & comparisons
~60

guides & comparisons

countries in directory
6

countries in directory

pages indexed
61

pages indexed

avg. search position
54 → 11

avg. search position

The problem

Legal and medical topics where mistakes hurt people, and user data that falls under GDPR special categories.

Stack

  • Next.js 16
  • Server Components
  • Server Actions
  • TypeScript
  • Tailwind
  • shadcn/ui
  • next-intl
  • Supabase RPC
  • Vercel Cron
  • Claude Code routines

Engineering decisions

  1. 01Privacy by design: no accounts until demand is measured — demand is tracked through tagged email signups instead.
  2. 02Supabase locked down: RLS on with no policies; the public key can only call two SECURITY DEFINER functions and never read a row.
  3. 03Every claim checked against primary sources — caught a closed regulator and a destination that stopped accepting foreigners before publishing.
  4. 04Fully static: typed TypeScript content modules rendered by shared templates, no CMS, JS only in small islands.
  5. 05Twice-daily Vercel Cron heartbeat keeps the free database from auto-pausing.
  6. 06Intro video cut from 42 MB to 3.4 MB with faststart; images 100–300 KB.
  7. 07Directory link checker: a plain script checks every organisation in the directory and reports dead links. A 403 counts as bot protection, not a broken link.
  8. 08Scheduled AI agent for legal accuracy: a Claude Code routine runs twice a week. It runs the link check and scans for changes in Dutch family law, with two independent sources for every fact. It is read-only and only leaves a draft report for human review. It never edits or publishes.

Outcome — A credible, fast MVP that validates demand before taking on the responsibility of storing sensitive data.

04 / Shopify e-commerce

Live2026

Sokole

A bilingual store for a Greek niche-fragrance brand selling to Greece and the EU. Built end to end: platform choice, design, theme, localisation, markets and handover.

sokole.eu
Sokole storefront hero, English version
Sokole mobile header and hero
Sokole philosophy section and signature fragrances
design iterations
2

design iterations

languages EN / GR
2

languages EN / GR

countries shipped
27

countries shipped

products launched
16

products launched

The problem

A small brand with a luxury product, a Greek and an EU audience, and Greek payment rules that most gateways don't support.

Stack

  • Shopify OS 2.0
  • Liquid
  • Sense theme
  • CSS custom properties
  • Shopify CLI
  • Shopify Markets
  • Translate & Adapt

Engineering decisions

  1. 01Chose Shopify over WooCommerce from the client briefs — lower maintenance, built-in Markets and payments — and walked the client through why.
  2. 02Delivered a second, more luxurious design direction after the first review.
  3. 03Custom sections: split editorial hero, olfactory-pyramid notes, product showcase, trust bar — on a CSS-token design system.
  4. 04All hardcoded theme strings moved to locale keys; Greek content bulk-updated through a CSV export → merge → import pipeline.
  5. 05Researched the Greek IRIS payment mandate: found the bank's gateway had no Shopify integration and proposed a working provider combination.
  6. 06Git-versioned theme with diff-before-push to live and theme-check.
  7. 07Handoff lesson: I built the store as a Shopify Development store, and that type can't be transferred to a client. In one working session I moved it to a store the client owns. The theme went over via CLI and 15 products via CSV. Collections, menus, markets and shipping were rebuilt. Translation CSVs are keyed by store-specific IDs, so a small Python script re-mapped 100 Greek strings by matching their English source text. The domain moved with no DNS changes. It's now a written playbook: a Client transfer store from day one, English as the default language, and ownership transferred only after full payment.
  8. 08Launch-day fixes: the catalog showed a placeholder T-shirt instead of the gift card and the footer "Gift Cards" link returned 404, because the product handle changed during the move to the client's store. I restored the gift-card handle and added a redirect from the old one. The newsletter heading read as if a fragrance were out of stock, so I rewrote it with the client in English and Greek.

Outcome — Launched in November 2026 at sokole.eu in English and Greek and handed over to the client, who now runs the store.

05 / Visual storytelling · AI video

Published2026

Don't Develop

A noir techno-thriller I write and publish on this site, in Russian and English. I also turn it into cinematic video: the full-length Prologue episode on YouTube is made entirely by me — script, images, voices, captions and edit.

Watch the Prologue on YouTube
Don't Develop Prologue video thumbnail
Don't Develop English book cover
Не проявляй Russian book cover
chapters written
47

chapters written

languages RU / EN
2

languages RU / EN

visual beats in the Prologue
51

visual beats in the Prologue

full-length episode on YouTube
1

full-length episode on YouTube

The problem

Turning a novel into video on a zero budget, while keeping the same faces, the same night-time London and readable on-screen text from shot to shot.

Stack

  • Stable Diffusion XL
  • Draw Things
  • Neural TTS
  • ffmpeg
  • Python
  • Next.js 16
  • Vercel Blob
  • Claude Code

Engineering decisions

  1. 01Script to beats: the Prologue is split into 51 visual beats and 92 spoken lines in one JSON file. Each beat names the speaker, the voice and the image.
  2. 02A visual canon: a style bible (photoreal noir, teal and amber, night, rain) and reference images for every character. A generated shot is rejected if a face drifts from the reference or a night scene comes out in daylight.
  3. 03Images generated locally with Stable Diffusion XL in Draw Things, with no subscription. Renders run one at a time, because parallel requests piled up behind slow ones and timed out.
  4. 04A separate neural voice for each character. The AI assistant Clara deliberately gets a non-British voice, so she sounds out of place even with the sound off.
  5. 05Captions are timed from the generated audio clips themselves, so no speech recognition is needed. On-screen text is added as an overlay, never drawn by the image model, which garbles it.
  6. 06The reading site is a Next.js app with chapters from Markdown and JSON, scene images on Vercel Blob and a video link for each chapter.

Outcome — A full-length Prologue episode on YouTube and a pipeline that works for any chapter. It's the same skill set I bring to stores: product imagery, hero videos and launch visuals.

Freelance

What I can build for you.

Fixed-price projects, remote, for clients across the EU. You get a written scope before work starts and a store or site you fully own at the end.

  • Shopify store, end to end

    From platform setup to launch: a custom theme, products, markets and shipping, a second language, payment research and a clean handover to an account you own.

    See: Sokole

  • Shopify fixes & migrations

    Theme changes, translations, redirects and broken links, speed, and moving a store between accounts or from WooCommerce without losing content.

    See: Sokole

  • Next.js sites & MVPs

    Fast, multilingual, search-ready sites and MVPs on Vercel and Supabase, built with the running costs in mind.

    See: LiveCity · TravelHub · FamilyFound

  • Product visuals & video

    AI-assisted imagery and video for stores and launches: hero loops, product scenes and short promos, with a consistent look from shot to shot.

    See: Don't Develop

Send a short brief →

Engineering notes

What production taught me.

Cost

Prefetch is not free

Unthrottled Link prefetch on camera cards triggered mass ISR writes, and routes without generateStaticParams were silently dynamic. Prefetch off, longer revalidation, English-only prebuild: daily Vercel usage down ~35%.

Bots

The scraper that drained the database

A crawler spike burned through the Supabase IO budget, saturated the connection pool and pushed functions into 300 s timeouts. Lookups cached, and one scraper — 71% of all requests — blocked at the Vercel firewall.

Limits

Zero cameras, no errors

At ~600 cameras the cached catalogue crossed the 2 MB data-cache entry limit and production quietly rendered an empty list. Reads are now chunked at 200 rows per cache entry.

Handoff

The store that couldn't be handed over

I built a client's Shopify store as a Development store. At handoff it turned out that type can't be transferred, can't take a paid plan and can't drop its password. In one session I moved everything to a store the client owns: theme via CLI, products via CSV, Greek translations re-mapped by a script, domain switched without touching DNS. Every client store now starts as a Client transfer store.

SEO

Thin pages are a liability

~1,200 untranslated duplicates were diluting a young domain. Unprocessed pages went noindex, untranslated locales 308 to English, and a content-only locale lets one article exist in one language.

Data

Don't trust a cron that writes

An automated liveness job flagged 33 working streams as offline. Rolled back: checks now produce a report, and a human approves every write.

Experience

Writing code since before AI wrote it.

PHP, Drupal, WordPress and Symfony in Dutch agencies from 2014 — the public repositories from those years are still on GitHub.

  1. 2026 – now

    Freelance Developer · Client work & own products

    Remote

    The Sokole Shopify store for a client; LiveCity.cam, TravelHub.cam and FamilyFound as my own products; AI video for the novel Don't Develop. Next.js, TypeScript, Supabase, Vercel, Shopify. AI-assisted engineering with OpenAI and Claude Code.

  2. 2018 – 2025

    Independent projects · Freelance

    Remote

    Rebuilt and maintained multilingual WordPress sites, including dutchnews.ru and fastgreece.

  3. 2017 – 2018

    Frontend Developer · WEBSOLVE B.V.

    Netherlands

    Portals and websites on a custom PHP framework and Symfony 3; responsive layouts and HTML email templates, from design to production.

  4. 2016 – 2017

    Software Engineer (Frontend) · Marviq BV

    Netherlands

    Frontend for Drupal projects with SASS, JavaScript and PHP; cross-browser implementation of UI designs.

  5. 2016

    Drupal Developer · Atom Group B.V.

    Netherlands

    Drupal 7/8 sub-themes, e-commerce and custom modules.

  6. 2015

    WordPress Developer (Intern) · KIIROO

    Netherlands

    WordPress development, customisation and frontend maintenance.

  7. 2014

    PHP Programmer (Intern) · Zorgcampus

    Netherlands

    PHP backend development and object-oriented fundamentals.

Education — Brunel University London · Design & Branding Strategy (MA programme) · 2006–2007. Omsk State Technical University · Engineering, printing & design · 1994–1999.

Stack

Tools I use in production.

Frontend

  • React 19
  • Next.js 16 App Router
  • TypeScript
  • Server Components
  • Tailwind
  • Framer Motion
  • CSS architecture / BEM

Backend & data

  • Supabase / Postgres
  • Row-level security
  • Server Actions
  • REST APIs
  • PHP
  • Symfony
  • Python scripts

Platform & cost

  • Vercel ISR & data cache
  • Cache invalidation
  • Firewall & bot control
  • Cron jobs
  • Usage monitoring

Search

  • i18n & hreflang (13 locales)
  • Structured data
  • Sitemaps & IndexNow
  • Search Console
  • AI search visibility

Commerce & CMS

  • Shopify OS 2.0 / Liquid
  • Shopify Markets
  • WordPress
  • Drupal 7/8

AI workflow

  • OpenAI — since 2025
  • Claude Code — since 2026
  • Anthropic API
  • Human-reviewed AI pipelines

About

Crutches, and then wings.

I started as a PHP and Drupal developer in Amsterdam agencies in 2014, after studying design and printing engineering. Live webcams were a hobby — I could watch skylines for hours.

In February 2026 I started turning that hobby into LiveCity.cam, learning React and Next.js as I built it — first with ChatGPT, from May with Claude Code. What would once have taken me months, or never shipped at all, went live in weeks. A year later there are four products in production, and I understand what sits underneath them: load, cache, bots, cost and search.

Available for freelance projects · Shopify & Next.js · EU citizen, remote

Working in five languages

  • English
  • Ελληνικά
  • Русский
  • Nederlands
  • Latviešu

Get in touch

Let's talk.

Have a Shopify store or a Next.js project in mind? Send a short brief — I reply within 24 hours with questions or a fixed-price quote.