Client: Iconic Group Inc. — a Canadian digital agency serving regulated, enterprise clients
Engagement: Multi-year contract product engineering (ongoing)
Role: Senior developer embedded in the product team
Stack: WordPress multisite, PHP, TypeScript/JavaScript, Blade/Twig, Docker, Playwright, GitHub Actions
Some of the most rewarding work happens inside long engagements, where you stop being a vendor delivering a fixed scope and become part of a team building a product over time. For the past year I have worked as a contract engineer with Iconic Group, a digital agency that builds and operates websites for large, regulated organisations.
This is a general account of the kind of work I do there. Out of respect for a mutual confidentiality agreement, it describes categories of work, technologies, and engineering practices — not client specifics, internal architecture, or anything proprietary. The goal is to give a sense of the shape of the engagement.
The Context
Iconic builds on a shared WordPress multisite platform: a common foundation of containerised environments, a parent-and-child theme hierarchy, a core layer of must-use plugins, and command-line tooling that ties it together. Many client sites run on one well-maintained base rather than a sprawl of one-off builds. That model is powerful, but it raises the bar: anything you add has to be robust enough to serve a fleet, and disciplined enough not to destabilise it.
My work over the past year has spanned four connected workstreams, all sitting on that shared foundation and all wrapped in an automated quality layer.
A Modular Page-Building System
A large share of the work involved a block-based page-building system: a library of reusable, configurable content sections that non-technical editors can assemble into pages without touching code. The engineering challenge in this kind of tooling is rarely the individual block — it is keeping a growing catalogue consistent, making blocks composable, and letting content move cleanly between environments through import and export. I contributed blocks, refined the editing experience, and hardened the plumbing that keeps the whole system coherent as it grows.
A Form Builder Built for Real Data
The largest workstream was a form-building system — the kind of tool that looks simple from the outside and is anything but. Forms are where a website meets real people and real data, so the hard parts live below the surface: conditional logic that shows and hides fields based on prior answers; server-side validation you can trust regardless of what the browser does; a submission pipeline with defined states; file uploads; notifications; a clean REST interface; and outbound delivery to downstream services via webhooks.
Because these forms handle personal information for regulated clients, the sensitive parts matter enormously: encryption of submitted data at rest, layered spam protection, and a storage model with sane retention. I built and refactored a great deal of this system across successive phases, moving it from an early prototype to a tested, config-driven architecture where new field types and processing steps are added through configuration rather than scattered special cases.
A Brand-Asset Generation Pipeline
A different flavour of problem: generating brand assets — logos and lockups — programmatically, in multiple formats and layouts, from a single source. This is the kind of work where pixels genuinely matter. Vector and raster outputs have to stay perfectly aligned; promotional and minimal variants have to scale without clipping or drift; and multiple language versions have to come out of the same pipeline. Much of my time here went into resolving sizing and offset discrepancies so that every generated variant was correct and consistent, then wrapping the generator in tests so it stayed that way.
An AI-Assisted Delivery Harness
The fourth workstream cuts across the other three. A meaningful part of this work is done with AI coding agents — but never as a free-for-all. I use a constrained harness that puts an agent inside a strict, gated workflow: it has to load the relevant documentation and specifications before it can touch anything; it works against approved plans; it writes tests before implementation; and hard gates block edits, pushes, and deploys unless the preconditions are met. The agent moves fast, but only inside boundaries it cannot cross.
That harness is how a single engineer can sustain the pace this engagement asks for without sacrificing the discipline that regulated, fleet-wide software demands. I have written about the workflow itself separately; here it is enough to say that it turns AI assistance from a novelty into a dependable part of the delivery process.
The Engineering Practices That Hold It Together
Across every workstream, the constant is quality infrastructure. The engagement runs on behaviour-driven specifications written in plain-language Gherkin, backed by end-to-end tests that exercise the software the way a user would. Builds run in containers so every developer and every CI run works from the same environment. Formatting and linting are automated. New capability arrives with its tests attached, and coverage is audited rather than assumed.
This is the part of professional software work that clients rarely see and always benefit from. It is why a form that handles personal data can be changed with confidence, and why a page-building system can grow for a year without collapsing under its own weight.
What This Kind of Engagement Delivers
A year of embedded product engineering with Iconic Group has produced, in general terms:
- A form-building system taken from prototype to a tested, config-driven architecture — with conditional logic, trustworthy validation, encryption at rest, spam protection, a REST interface, and webhook delivery
- A modular page-building system extended with new blocks, a smoother editing experience, and reliable content import and export
- A brand-asset generation pipeline producing pixel-accurate, multi-format, multi-language outputs, protected by tests
- A repeatable, test-driven workflow for delivering that work with AI assistance safely and at pace
- A large, maintained suite of behaviour specs and end-to-end tests that make ongoing change safe on a fleet-wide platform
Why It Works
Long engagements reward a particular temperament: a willingness to fix the source rather than patch the output, to invest in tests and tooling before they are strictly needed, and to treat someone else’s platform with the same care you would give your own. That is the way I like to work, and it is why this collaboration with Iconic Group has continued to grow.
If your organisation is looking for an engineer who can embed in a team and deliver production software under real constraints, that is exactly the kind of work I do.