Delivery in practice
Case Studies
Working software, usable technical knowledge and outcomes you can inspect. See what AI contributed, where human judgement mattered and how each result was checked.
Theme
Subject
Showing 10 of 10 case studies.
AWS platform engineering and security governance
AWS security guardrails that keep delivery moving
An oil and gas company turned AWS security policy into a reusable delivery platform, with workload-specific controls and a tested path to tighter boundaries.
Project work ·
Outcome
A repeatable policy delivery platform, extended with an internet-exposure pilot validated against 90 days of activity across 62 sampled accounts.
AI delivery tooling
Local models, frontier judgement — building a practical AI delivery workflow
How experiments with local language models became a working delivery toolchain, combining measured performance, isolated coding tasks and independent review.
Outcome
A working local delegation toolchain, with measured model trade-offs and a clearer boundary between useful automation and work requiring review.
AI-assisted software delivery
From WordPress export to an AI-editable website
We rebuilt our live marketing site around structured source, using AI-assisted delivery and explicit checks to carry its content and established links forward.
Outcome
A live site maintained directly in source, with tested publishing and legacy URL coverage.
Technical knowledge and applied learning
From scanned manuals to a working restoration
An AI-built technical resource helped turn specialist documentation into usable knowledge and supported a full recap and calibration of an Otari MX-5050.
Outcome
A technical reference website and a completed full recap and calibration.
Energy systems
Two battery systems, one energy plan — built through LLM conversations
How I used LLM conversations to build a working energy controller, connect FoxESS and EcoFlow systems, and learn to coordinate solar production, battery storage and wholesale electricity prices.
Outcome
$4.16 credit against energy usage costs over the first 19 days with Amber, consuming around 40–45 kWh/day.
AWS architecture, security and AI-assisted delivery
Beyond shared networking: an AWS sandbox for engineering
A sandbox-owned network and local Transit Gateway gave an oil and gas company a foundation for AI-assisted engineering and a path to Amazon WorkSpaces.
Project work ·
Outcome
An EC2 sandbox prototype with its own network and Transit Gateway, plus groundwork for a future Amazon WorkSpaces platform.
Code security tooling assessment
Assessing code security tools for domain specialists’ work
Assessing security scanning for an oil and gas company’s specialist analysis, automation and B2B services, using estate measurements, local scanner adapters and sourced AI research to prepare a reviewable decision.
Project work ·
Outcome
An assessment foundation for evaluating security-tool fit: common estate and scan inputs, shortlist generation, report and deck outputs, and dashboard preparation.
Enterprise AI release engineering
An enterprise AI application needs a delivery system
Principal engineering for an AI application’s delivery pipeline: local feedback, blocking checks, policy tests and usable security results, with release acceptance kept explicit.
Project work ·
Outcome
A versioned delivery-control foundation for an existing AI project; production release and application acceptance remained separate milestones.
AWS security architecture and platform delivery
Retiring a US$43,000 WAF contract with a shared AWS platform
Replacing Imperva with a shared AWS security platform removed an annual renewal and made application onboarding a reviewed configuration change.
Project work ·
Outcome
Retired a US$43,266 annual Imperva contract; 21 application distributions use four shared ALBs with configuration-driven onboarding.
AWS platform engineering and service automation
EC2 power leasing across 13 AWS accounts
Growing account footprint; about US$699,000 in cumulative on-demand-equivalent compute value.
Project work ·
Outcome
Cumulative on-demand-equivalent compute value: approximately US$699,000 · Savings period: 1 March 2025–30 September 2026 (19 months) · Annual estimate: approximately US$360,000.
What could this approach deliver for your organisation?
Start with an Enterprise AI Delivery Assessment: a candidate use case, clear acceptance criteria and a bounded delivery plan.
Explore the assessment