§ WORK
Built, shipped, measured.
My own products, client work and internal tooling. Pick a project to explore it.
- years of production PHP
- 10+
- client ecommerce
- 5 yrs
- shops in a live catalogue
- 14
- signed data-licensing agreement
- IGDB
Kiwi Game Bazaar
FIG. W-01 · THE PIPELINE AS A DISTRICT
DATA FLOW
- 14 SHOPSnightly feeds
- ADAPTERSone per store
- RAWappend-only
- CANONICALconfidence-scored
- MARKETPLACEkiwigamebazaar.co.nz
Claude wrote most of the code under my direction; every merge and deploy needed my approval
- shops, nightly
- 14
- numbered decisions
- 320+
- test code to app code
- ~1:1
- signed licence
- IGDB
THE BRIEF
NZ’s physical games were spread across niche stores, with nowhere to compare them. As a gamer, I built the place that didn’t exist.
WHAT I BUILT
A reference database and live-availability marketplace: an ingestion pipeline that keeps the raw data behind every record, 14 shop integrations, pg_trgm matching, SEO infrastructure and a compliance framework over every integration.
OUTCOME
Live, with nightly prices and stock across 14 shops, a signed IGDB licence, and a test suite roughly the size of the codebase.

Screenshot of kiwigamebazaar.co.nz.
Games metadata is powered by IGDB.com: https://www.igdb.com
Live tournament platform
FIG. W-02 · THE NIGHT AS A DISTRICT
DATA FLOW
- QR SIGNUPat the door
- SNAKE DRAFTskill-weighted
- REVERBlive broadcast
- SCREENSbracket display
- ADMINanalytics + MCP
- on Laravel Cloud
- Live
- automated security audit
- Weekly
- agent-ready admin
- MCP
THE BRIEF
A live weekly gaming event needed fast signups, fair teams and a bracket the whole room could follow.
WHAT I BUILT
A real-time tournament manager: QR signup, a skill-weighted snake-draft algorithm, live bracket display, an admin portal with analytics, and admin actions exposed to AI agents over MCP.
OUTCOME
Deployed on Laravel Cloud with a custom domain, full Pest and Vitest coverage, and a weekly automated security audit.

Recorded on demo data: every name is a made-up gamertag.
Client ecommerce, wired in
FIG. W-03 · THE STACK AS A DISTRICT
DATA FLOW
- ERP / POSinventory + orders
- INTEGRATIONclient-specific
- STOREFRONTclient ecommerce
- MARKETINGDotdigital · Bloomreach · Mailjet
- REPORTINGGA · Meta
- client ecommerce
- 5 yrs
- marketing platforms
- 3
- repeat incidents after the fix
- 0
THE BRIEF
Client ecommerce sites that had to stay in step with ERP, POS, inventory and marketing systems, each set up a little differently.
WHAT I BUILT
ERP/POS and inventory integrations, Dotdigital, Bloomreach and Mailjet integrations, legacy-site migrations, and ecommerce and SEO reporting via GA and Meta. Led code reviews, testing and planning across the team.
OUTCOME
Migrations rehearsed and reconciled before cutover, a production incident I owned and fixed with no recurrence, and a team process change to confirm real field names before building.
Command Centre
FIG. W-04 · THE AI SETUP AS A DISTRICT
DATA FLOW
- DASHBOARDlaunches a run
- HORIZONqueued as a job
- HEADLESS RUNClaude Code
- REVERBstreamed live
- DASHBOARDresults + history
- results over Reverb
- Live
THE BRIEF
One place to manage my Claude Code setup: its skills, scheduled routines and memory.
WHAT I BUILT
A Laravel/Inertia dashboard that launches headless AI runs through Horizon queues and streams the results back over Reverb.
OUTCOME
AI runs started and watched from one screen instead of a pile of terminals, with results streamed back live.
Music Bazaar (working name)
FIG. W-05 · THE CATALOGUE AS A DISTRICT
DATA FLOW
- SHOPSasked first
- ADAPTERSone per shop
- BARCODE MATCHto one album
- ALBUM CATALOGUEDiscogs CC0 dumps
- ALBUM PAGESnot yet public
Every shop was asked before any of its stock was read
- barcoded CDs matched one album
- 87%
- album catalogue
- CC0
- asked before reading
- Consent
THE BRIEF
Used CDs and cassettes sit in NZ shops with nowhere to see who has an album in stock.
WHAT I BUILT
A sibling site on Kiwi Game Bazaar’s ingestion engine, matching each shop’s used listings to one album by barcode against the Discogs CC0 data dumps.
OUTCOME
87% of a real shop’s barcoded CDs matched one album. Shops were asked before any stock was read, and the conditions for stopping were written before building. Not yet public.
3D Weather Globe
FIG. W-06 · THE WEATHER AS A DISTRICT
DATA FLOW
- ERA5climate reanalysis
- PYTHONpandas pipeline
- LICENCE AUDITcommercial use
- WEBGPU GLOBEthree.js
- 39 PINSeach month’s main event
Day counts are withheld where the model disagrees with station records by 1.5× or more
- destination pins
- 39
- model-vs-station gap that withholds a count
- 1.5×
- three.js renderer
- WebGPU
THE BRIEF
Live weather globes and best-month-to-travel maps both exist, but nothing animates the patterns month by month with a zoom into destinations.
WHAT I BUILT
A browser globe built with Vite, Vue 3, TypeScript and three.js on WebGPU, over a Python/pandas pipeline of ERA5 climate data. The engine was chosen by a one-week Three.js vs Babylon.js spike against criteria written first.
OUTCOME
39 destination pins show each month’s main weather event. Every dataset is licence-audited for commercial use. In development, not yet public.

Wind, cloud, precipitation, temperature and model ground height: ERA5 (Copernicus Climate Change Service)
Generated using or contains modified Copernicus Climate Change Service information 2026. Neither the European Commission nor ECMWF is responsible for any use that may be made of the Copernicus information or data it contains.
Copernicus Climate Change Service (2023): ERA5 monthly averaged data on single levels from 1940 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS). DOI: 10.24381/cds.f17050d7 (Accessed on 01-Nov-2026, 02-Nov-2026 and 11-Nov-2026)
Hersbach, H., Bell, B., Berrisford, P., Biavati, G., Horányi, A., Muñoz Sabater, J., Nicolas, J., Peubey, C., Radu, R., Rozum, I., Schepers, D., Simmons, A., Soci, C., Dee, D., Thépaut, J-N. (2023): ERA5 monthly averaged data on single levels from 1940 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS), DOI: 10.24381/cds.f17050d7 (Accessed on 01-Nov-2026, 02-Nov-2026 and 11-Nov-2026)
Copernicus Climate Change Service (2025): ERA5 hourly time-series data on single levels from 1940 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS). DOI: 10.24381/1cf1ad76 (Accessed on 11-Nov-2026)
ERA5 data licensed under CC BY 4.0: https://creativecommons.org/licenses/by/4.0/
Cyclones: IBTrACS (NOAA NCEI)
NOAA's International Best Track Archive for Climate Stewardship (IBTrACS) data, accessed on 2 November 2026
The cyclone layer counts only the US agencies' best tracks in IBTrACS (NHC, CPHC and JTWC).
Earth imagery: NASA
Blue Marble: Next Generation: NASA Earth Observatory
No endorsement by NASA is implied.
§ AT A GLANCE · CLICK TO OPEN





