Technologies and architectures
What this page covers
The architecture, experience and data patterns behind the systems we build, where each one was proven, how technologies are chosen, and which partnerships support the work.
Every system we build starts from solutions already running somewhere else. A pattern born on one application is written down, tested again and carried to the next. Choosing a technology is a decision that lasts years, so every choice carries a reason that can be examined.
Patterns that have already run in production
A pattern is a way of solving a problem: how a platform rebuilds itself from code, how a design system reaches the applications that use it. Each appeared first on a single application under real load: the platform behind our own products measured 99.991% availability over a year. We write it down in a form another team can pick up, and the next project starts there. Work done on one system then improves the others.
Go deeper
The patterns that travel
Four patterns recur across industries, and each runs today in systems we built.
Working without a connection
Applications keep a local copy of the data they need and carry on while the network is down. When it returns, the device sends only what changed, and the people using it never stop working.
Data that stays in step
When several devices change the same record, the architecture decides how the conflict is resolved. Across a network of more than 200 sites, twelve million stock records come into line in 200 seconds.
One platform, separate brands
Several brands or countries share one platform while their data, users and permissions stay apart. Each keeps its own look, and a new one joins without a rebuild.
Models that stay accurate
An AI model in daily use drifts as the data changes. Versions, quality indicators, held-back test cases and a route back to the previous version keep it reliable.
The pattern moves, the code stays
Reuse across companies raises a question about ownership, answered before the work starts. What moves from one project to the next is the way a problem was solved: the architecture, the order of the steps, the rules we learned to apply. Software written for a company belongs to that company, with its infrastructure code and its documentation. A pattern makes the second build faster; it never makes two systems the same.
Go deeper
Where the engineering goes deep
Four areas recur in the systems we build, each with the tools that have carried work in production.
Cloud and DevOps
Kubernetes clusters described entirely in code and realigned with the repository by an automated check. We work on Azure, AWS and Google Cloud, alone, combined or alongside in-house servers.
Go deeper
AI and data
Language models for assistants and agents, retrieval over a company's own documents, and networks trained for one repeating decision with TensorFlow or PyTorch. Quality and cost per request are followed after release.
Go deeper
Edge and offline-first
A database on the device, selective synchronisation and a gateway that reconciles it with the central system, so work continues on site when the connection drops.
Design systems and product platforms
Components built once and installed by every team, in React, Angular or Vue, with design tokens held in one source. Independent services talk through interfaces and events, and product data flows from a PIM to every channel.
How a technology is chosen
Five things decide a choice. The technology has to meet the constraints of the case and be the simplest option that does. The in-house team must be able to maintain it, and its cost is counted across its whole life. One part has to be replaceable without the rest. Where an agreement includes reselling licences, the technology still has to meet the same conditions. We call a technology proven only when it runs at a company or inside our own products.
What our partnerships make possible
A partnership counts for what it makes possible on a project. The agreement with Couchbase covers the database that runs on devices in the field and the gateway that keeps it in step with the central system, and it opens a line to the vendor's architects on an unusual design. As a certified Shopify Partner, we move shop tills onto the same platform as the online shop. An agreement with Combeenation supports the rules behind a product configurator.
Contact us to validate the technical feasibility,
business impact and build your business case together.