Custom software development
Bespoke web and desktop applications built around how your organisation actually works, rather than bending your process to fit a package.
PythonJavaTypeScriptPostgreSQL
Services
Everything from the first line of code to the first page of Google.
22 services across four practices. Most clients start with one and end up using three, because the reason a product is not working is rarely where they first thought it was.
Software development
Systems people use every day, built to survive contact with real users.
Each of these has shipped for a real client or into one of our own products. We do not list a capability we have not built with.
SaaS and multi-tenant platforms
Subscription products with the parts nobody demos: tenant isolation, role-based access, audit trails, billing hooks and a migration path.
ReactFastAPIMulti-tenancyRBAC
Mobile app development
iOS and Android applications, including on-device inference and private distribution where an app is for one organisation rather than the public store.
iOSAndroidTensorFlow LiteApp Store Connect
Web development and e-commerce
Marketing sites, portals and storefronts that load fast, rank well and hold up on a mid-range phone, not just on your designer's laptop.
Next.jsStatic sitesCMSCore Web Vitals
API development and integration
Connecting the systems you already pay for. Microsoft Graph, finance systems, student records, CRMs and anything with an endpoint.
RESTGraphQLMicrosoft GraphWebhooks
Legacy modernisation
Moving an ageing system forward without a rewrite-everything gamble. We have taken desktop Java estates to modern interfaces in place.
RefactoringStrangler patternData migration
Artificial intelligence and data
Models trained on your data, evaluated honestly, and shipped into something someone uses.
Each of these has shipped for a real client or into one of our own products. We do not list a capability we have not built with.
AI model training and fine-tuning
Custom models trained on your own data, from dataset construction and labelling through training, evaluation and deployment. Including fine-tuning open models where a general one will not do.
PyTorchTensorFlowFine-tuningModel evaluation
Machine learning and prediction
Classification, ranking, forecasting and risk scoring, measured against a simple baseline first so you know what the sophistication actually bought.
scikit-learnPandasFeature engineering
Computer vision
Detection, segmentation and automated quality inspection, including on-device inference where there is no reliable connection.
Mask R-CNNSegmentationEdge inference
LLM applications and AI agents
Retrieval-augmented assistants, document understanding and task agents grounded in your own content, so answers can be checked against a source.
RAGVector searchOllamaOpenAI API
Data engineering and analytics
Pipelines, warehousing and dashboards that turn scattered operational data into something a manager can act on this week.
ETLPostgreSQLDashboardsReporting
AI workflow automation
The repetitive work in your organisation, automated end to end: document processing, data entry, report generation, triage and routing.
AutomationDocument AISchedulingIntegrations
Growth and design
Being good is not the same as being found. Both are engineering problems.
Each of these has shipped for a real client or into one of our own products. We do not list a capability we have not built with.
SEO and technical optimisation
Technical audits, site architecture, structured data, Core Web Vitals and content strategy. We fix the crawlable foundations before touching keywords.
Technical SEOSchema markupPage speedContent
Digital marketing and content automation
Campaign systems that plan, generate and publish on schedule through official APIs, so growth does not depend on someone remembering to post.
Marketing APIsSchedulingAnalyticsCopy systems
UI and UX design
Interface design and design systems, including the unglamorous parts: empty states, error messages, accessibility and keyboard paths.
Design systemsPrototypingWCAGFigma
Brand and identity systems
Logo, palette, typography and asset libraries built as a system your team can apply without coming back to us for every graphic.
IdentityAsset pipelinesTemplates
Infrastructure and support
Getting it live is half the job. Keeping it live is the other half.
Each of these has shipped for a real client or into one of our own products. We do not list a capability we have not built with.
Cloud, DevOps and hosting
Deployment pipelines, environment management, monitoring and cost control on Azure and AWS, so releases stop being events.
AzureAWSDockerCI/CD
IoT and embedded systems
Sensors, firmware, telemetry and the application on the other end. Complete loops from a reading in the field to an action taken.
ESP8266FirmwareFirebaseTelemetry
QA, testing and security review
Functional and end-to-end testing, API security auditing and penetration-style review of the endpoints your application exposes.
Test automationAPI securityAudit
IT infrastructure and migrations
Microsoft 365 tenancy work, domain and identity migrations, and virtual learning environment moves, planned so no historic data is lost.
Microsoft 365Entra IDMoodlePowerShell
Support and managed services
Ongoing maintenance, monitoring and enhancement under a plain agreement, so you are never locked in by opacity.
SLA supportMonitoringEnhancements
Training and capability building
Practical upskilling for your team on AI tooling, engineering practice and data literacy, delivered by people who teach for a living.
WorkshopsCurriculumMentoring
How we work
Five stages. The first two are where most projects should change shape.
We would rather narrow a project in week two than discover in month five that the data never supported the claim.
Frame the decision
Before a model, the decision. Who acts on the output, what changes when they do, and what the cost of a wrong call actually is.
Audit the data
What exists, what is missing, what is biased and what cannot legally be used. Most projects narrow here, and that is the point.
Baseline first
A simple, cheap baseline gets built and measured before anything sophisticated. If the baseline is good enough, we say so.
Build the explanation with the model
The output and the reason for it are one deliverable. A score nobody can interrogate is a liability, not a feature.
Ship, watch, correct
Into production behind real access controls, with monitoring, an audit trail and a defined route for a person to challenge a decision.
Engagement models
Three ways in, depending on how much certainty you already have.
If you are not sure which fits, describe the problem and we will tell you which one we would propose, including the smallest one.
Discovery sprint
Two to three weeks. We work through the data you hold, what a model could realistically add, and what it would cost to run. You end with a decision, an architecture and a cost model — including the option to stop.
2–3 weeksFixed feeWritten recommendation
Build partnership
We design and build the system, then hand it over with the documentation and tests to run it yourself. Delivered in reviewable increments, with production access and a support window at the end.
8+ weeksMilestone-basedHandover included
Embedded AI team
Our engineers work inside your team on your roadmap, for the period you need the capability rather than the headcount. Suited to organisations building a first AI function.
MonthlyNamed engineersYour tooling
Questions
Answers to what clients ask before signing anything.
If yours is not here, send it over. We answer directly rather than routing you to a form.
No. Education is where our deepest domain knowledge sits, and it is the hardest test of explainability because a decision affects a person's record. The same engineering applies to inspection, agriculture, operations and any setting where a model informs a decision someone has to justify.
You do, unless we agree otherwise in writing. Under our standard terms, all deliverables produced for a client vest in the client on payment. Where a project builds on our existing platform components, those stay ours and you get a licence — which we set out plainly before work starts.
Yes. We build under UK GDPR by default: data minimisation, defined retention, role-based access, audit logging and tenant isolation. For education clients we are used to working alongside institutional data protection officers and existing information governance policies.
We tell you. A discovery sprint that ends with a recommendation not to build is a successful sprint — you paid for weeks of investigation instead of months of construction. A rules engine or a better report is sometimes the honest answer.
Systems ship to your infrastructure or to managed hosting we run for you. Handover includes documentation, tests and a support window. Ongoing support is a separate agreement so you are never locked in by opacity.
Start with the smallest useful piece of work.
A discovery sprint costs weeks, not months, and ends with a written recommendation you can act on either way.