Work

Things we have actually built

A short list, honestly described. No invented metrics — just what the job was and what was delivered.

Case study · Hospitality & retail

Elwood Bakery

The brief. A bakery selling in person only, losing orders every time the doors shut.

What was built. A WordPress and WooCommerce store on a custom Astra child theme, with Stripe checkout and a mobile-first ordering flow. Built and tested locally in Docker, deployed to managed hosting, with a documented handover so the team can run it themselves.

WordPress 6.5 WooCommerce Stripe Custom child theme ACF
Visit elwoodbakery.com.au
Elwood Bakery home page, showing the catering hero banner

Product · Hospitality automation

Rest-O-Pi OS

The problem. Independent kitchens either pay enterprise prices for a POS or run on paper dockets re-typed at close.

What it does. A Raspberry Pi on the shop's own network receives Shopify order webhooks, prints the kitchen ticket automatically and drives a live kitchen display — and keeps working through an internet dropout.

Shopify webhooks Raspberry Pi 5 ESC/POS printing Kitchen display
See the full product page
Rest-O-Pi OS hardware on a kitchen counter — Raspberry Pi, thermal printer with a printed kitchen ticket, and a tablet showing the live order board

Case study · AI, data & automation

Turning noisy logs into triaged tickets — and a draft fix

The brief. A support team was spending its day on the mechanical half of incident response: reading log output across several tools, working out which spike actually mattered, raising the ticket, writing up what happened, then hunting for where in the codebase to start. Client name withheld at their request.

What was built. A pipeline that watches the logs, correlates them against the host metrics, and hands the candidates to an AI model for root-cause analysis — then files the result everywhere it needs to go: the Jira ticket, a draft pull request on GitHub, Bitbucket or Mercurial, and a summary in the team's channels.

The team still decides what ships. The difference is that the triage, the write-up and a first attempt at the fix are waiting for them, instead of being the job.

CloudWatch Datadog Python MCP servers Jira GitHub & Bitbucket Slack & Teams
How we approach AI adoption
An operations workspace with three monitors showing dashboards, charts and a log stream

How it runs

  1. 1

    Watch

    Log streams from CloudWatch and Datadog, scanned for recurring patterns.

  2. 2

    Correlate

    Those patterns lined up against CPU and the other host metrics.

  3. 3

    Diagnose

    An AI model works the root cause and keeps the evidence attached.

  4. 4

    Act

    Filed everywhere it needs to go, before anyone opens the ticket.

    • Jira ticket via MCP
    • Draft PR
    • Slack & Teams

Tell us what's slowing you down.

A free, honest conversation. If the answer is "don't change anything", we'll tell you that too.