Real systems. Real sectors.
Confidential by design.
Our team has spent years inside software that matters: platforms for government agencies, payment and accounting systems, corporate web platforms and our own AI tools. Each project is described by what it does and what we did, never by whose it is.
AI systems we designed, built and run.
Two of our own AI products, used in real day-to-day delivery.
Agentic QA: AI that tests like a senior engineer, and proves it.
A complete AI-driven testing system. Agents read the real requirement, plan coverage, test the live application in a browser and write the report, while people approve every step that matters.
- Requirements into rulesAcceptance criteria are split into numbered, testable rules from the real spec.
- Coverage you can countProgress is a ratio of rules covered, never a vague "looks good".
- Predictions firstExpected results are written down before testing, so surprises are visible.
- Live browser testingEvery case runs in a real browser with steps, expected, actual and a screenshot.
- Reproduced or it isn't a bugUnconfirmed suspicions are ruled out on the record, not filed.
- Human approval gatesFive checkpoints where a person signs off before anything moves on.
- Self-checking reportsAudit scripts reject incomplete reports before anyone reads them.
- Release-level decisionsRolls tickets up into one ready / not-ready call for a release.
- Writes real automationTurns approved cases into tests in the project's own framework.
- Team-readyInstalls in one command for any project, with guided setup.
R1.1Saving a record shows a confirmationPASS · screenshotR2.3Private field hidden from other usersBUG · reproducedR3.2Suspected script injection in notesRULED OUTGate DReport audit passed, awaiting sign-offHUMAN GATEAI project-management assistant: the busywork, handled.
An assistant for project leads. It watches the board, reads each new client request alongside the actual codebase, and prepares the work a lead would otherwise do by hand, for the lead to approve.
- Reads every new requestPicks up client tickets as they arrive and queues them for analysis.
- Understands the codeChecks each request against the real codebase, read-only.
- Implementation notesDrafts where the change lives and how to approach it.
- Asks the right questionsSpots ambiguity early and drafts clarifying questions for the client.
- Daily digestEvery working morning: what moved, what's stuck, what needs a decision.
- Finds hidden bottlenecksFlags work parked in "almost done" states that charts count as finished.
- Sprint reportingDrafts the completed-work summary for sprint pages automatically.
- Meeting follow-upsLinks clarifications from client meetings back to the right ticket.
- Human in chargeNothing is posted or assigned without approval, with a one-switch off.
- Proven in a blind pilotTested on finished work it had never seen, with no false claims about the code.
Projects, by sector.
Each project is shown on its own, with what the system does, what we did and a short animation of its core flow.
🔒 Client names, products and business rules are confidential and never shownPublic safety & justice
Government agenciesGuided reporting with automatic document generation
A web platform that guides staff through a detailed report step by step, prompting for information they might otherwise leave out, then fills in and generates every required document from one set of answers.
- Step-by-step guided entryFree text, drop-downs and check boxes that ask the right next question.
- More complete reportsPrompts capture details people often forget to record.
- One entry, every formAll required documents are generated from the same answers.
- Rule-heavy logicRegulated timelines and record-keeping rules, modelled and tested.
- Built for scrutinyOutput that holds up to formal review, so accuracy is non-negotiable.
Our role: requirements analysis, story sizing, test strategy, functional, regression and performance testing, automation, release sign-off.
auto-filled from answersready
auto-filledready
auto-filledready
deadline trackeddeadline set
Specialist evaluation records, from exam to final report
A system for certified specialists who carry out a structured, multi-step assessment and produce a final expert report. We also verified its move from a legacy system to a modern architecture.
- Multi-step evaluation workflowEach phase captured in a fixed order, nothing skipped.
- Structured measurementsReadings and checks recorded as consistent, structured data.
- Guided decisionsA fixed protocol keeps every assessment consistent.
- Expert report outputDocumentation ready for formal review.
- Safe modernisationLegacy and new system compared screen by screen before switch-over.
Our role: quality lead, mapping every evaluation workflow, AI-assisted side-by-side comparison of the new system against the legacy one, defect reporting.
One electronic report, shared across agencies
A reporting system that replaces handwritten forms and duplicate paperwork. Staff complete one electronic report, and it reaches every partner organisation involved in the case, so the whole process moves faster.
- Paper to electronicMany forms and templates combined into one easy reporting interface.
- No duplicate entryInformation typed once is reused wherever it is needed.
- Multi-organisation workflowEvery partner organisation works from the same source.
- Sensitive case handlingBuilt for confidential, high-stakes reports.
- Shaped by usersUsability improved through direct feedback from end users.
Our role: test planning, test design, functional, regression, performance, smoke and sanity testing, automation suite, usability sessions with end users.
Single sign-on for a family of applications
Shared identity and supporting services that let users move between several connected applications with one secure login, plus the regression safety net that keeps every connected app working after each release.
- One login, many appsUsers sign in once and reach every connected application.
- Consistent access rulesRoles and permissions respected across all systems.
- Cross-platform regressionEvery release checked against every app that depends on it.
- Coordinated releasesChanges rolled out in step across products.
Our role: quality lead across all connected platforms, cross-system regression, release coordination and sign-off.
secure sessionverified
Hospitality, fintech & accounting
Cloud software for restaurant groupsInvoice capture from a phone photo
Restaurants send supplier invoices as photos from a mobile app, by email or upload. Line items are captured automatically and flow straight into prices, order guides and food-cost numbers.
- Snap, email or uploadEvery way a restaurant receives invoices is supported.
- Automatic line itemsProducts, quantities and prices captured without typing.
- Live cost updatesNew prices immediately update recipes and food costs.
- Approval flowManagers review and approve before invoices post.
Our role: test planning and design, mobile app testing on Android and iOS, approval-flow testing, automation upkeep, test-cycle management.
Supplier bill payments, synced to accounting
Restaurants pay supplier bills from one place by virtual cheque, virtual card, bank transfer or batch payment files, across multiple locations, with every payment synced back to their accounting system.
- Many payment methodsVirtual cheques, virtual cards, bank transfers and batch payment files.
- Multi-locationPay for one restaurant or a whole group at once.
- Automatic accounting syncEvery payment lands in the books without re-keying.
- Data you can trustAmounts and statuses validated at database level.
Our role: payment-flow testing for every method, multi-location scenarios, SQL data validation, automation suite maintenance.
Point-of-sale to accounting, every single day
Daily sales are pulled automatically from dozens of point-of-sale systems, and sales and invoice entries are posted to the restaurant's accounting package in its own chart-of-accounts format.
- Dozens of POS systemsSales collected automatically, whichever till a restaurant uses.
- Major accounting packagesEntries posted into the books the restaurant already uses.
- Account mappingEvery sale lands in the right ledger line.
- Built for volumeDaily imports tested under load.
Our role: integration and data-sync testing, mapping validation, regression and load testing of daily imports.
Inventory, ordering and recipe costing
Inventory counts, supplier ordering from order guides, recipes shared across branches, plate costs that update with live prices, and reports comparing what should have been used with what actually was.
- Order guidesPlace, process and track supplier orders in a few taps.
- Recipe managementTrack costs, set menu prices and share recipes across branches.
- Live plate costsCosts recalculate the moment supplier prices change.
- Usage reportsExpected versus actual usage reveals waste and over-portioning.
Our role: functional and regression testing of costing logic, counts, order guides and food-cost reports.
Life sciences & enterprise web
Corporate platformsCorporate platform for a biotech company
A multi-section corporate website for a research-driven biotechnology company, built on an enterprise content-management system so editors can publish science, pipeline and investor news on their own.
- Enterprise CMSEditors build and publish pages from reusable components.
- Science contentResearch areas and pipeline explained for a broad audience.
- Investor-readyNews and investor information kept accurate and current.
- Release qualityEvery new feature and fix verified before going live.
Our role: feature and fix verification, defect tracking, requirement clarification with developers and project managers.
Business web application for an international client
A client-facing business web application delivered for an international customer, where we owned day-to-day quality across every release.
- Release confidenceSmoke and sanity checks on every build.
- Full regressionExisting features re-checked before each release.
- Clear defect reportsEvery issue logged with steps and evidence.
- Fast feedbackClose loop with the development and project teams.
Our role: functional, regression, smoke and sanity testing, test-case design, defect reporting, requirement clarification with developers and managers.
Delivery & practice leadership
Team and company levelLeading delivery across several products
Running a cross-functional team across several products at once: priorities, estimation, sprint planning, release go/no-go and direct communication with an overseas client, after moving the work to agile.
- Agile transformationProjects moved to agile, with more predictable sprints.
- Estimation and planningScope sized with the team, priorities agreed with the client.
- Release go/no-goClear, evidence-based release decisions.
- Pressure-proof deliveryTime-critical releases shipped without client escalation.
Our role: delivery and quality lead, Scrum ceremonies, backlog grooming, client communication, release decisions.
An AI upskilling programme for a testing practice
A career-progression and AI-adoption plan for a whole testing practice, from trainee to lead, with the skills expected at each level, plus company-wide sessions on working with AI. Built as part of the company's centre-of-excellence group.
- Clear career ladderWhat each level, from trainee to lead, is expected to know.
- AI at every levelPractical AI skills woven into everyday testing work.
- Coaching and reviewsMonthly growth conversations on AI use, coverage and domain knowledge.
- Company-wide sessionsTeaching teams how to excel as testers in the AI era.
Our role: designed the programme, presented it to management and led company-wide sessions.
Our own products
Built and shipped end to endCross-border tax compliance for online stores
A plugin for a widely used open-source e-commerce platform that handles cross-border tax rules at checkout. Published on the official plugin directory, with a free core and a paid tier with licensing.
- Tax-number validationBusiness customers verified live at checkout.
- Correct tax, automaticallyExemptions and destination rates applied without manual work.
- Freemium modelFree core plus a paid tier with reporting and thresholds.
- Secure licensingPaid features unlock with a licence that re-checks itself.
- Tested architectureBusiness logic separated from the platform and unit-tested.
Our role: product design, development, testing, licensing, payments and launch, end to end.
AI quality toolkit, installable by any team
The agentic QA method above, packaged as an installable toolkit, so any engineering team can run the same structured, evidence-first test runs on its own projects.
- One-command installTeams add it to their tools in seconds.
- Guided setupConfigures itself for each project on first run.
- Project-agnosticWorks on any codebase, with no client details inside.
- Tested releasesAutomated tests verify every version.
Our role: designed, built, documented and maintained, with versioned releases.
Corporate website, engineered like a product
A fast, secure corporate website for a technology company: service pages, an enquiry form that reaches the team's inbox, and a built-in assistant that answers visitor questions without sending anything to a server. Designed, built and quality-tested to client standards in two browser engines.
- Top Lighthouse scores100 for SEO, accessibility and best practices.
- Strict securityHardened security headers, no cookies, no third-party trackers.
- Private assistantAnswers questions in the browser, nothing leaves the device.
- Automated checksOver a thousand checks run on every change.
Our role: design, build, testing, security hardening and launch.
Personal builds
Side projects and experimentsTax-return filing assistant for a government portal
A personal AI agent that fills in an online income-tax return on a government e-filing portal from prepared data, field by field, and stops before anything is paid or submitted, so a person always makes the final call.
- Portal navigationMoves through a complex government web portal like a careful user.
- Data from one sourceFills every section from a single prepared set of figures.
- Draft onlyNever pays or submits; it stops for human review.
- Separate tax logicCalculations kept apart from form-filling, so each can be checked.
Our role: designed and built it for personal use.
A team of research agents that find real opportunities
A multi-agent system that scans online marketplaces for unmet demand, verifies every claim at its original source, writes a short dossier, and passes only evidence-backed ideas to a final reviewer agent and then to a person.
- Scout agentsEach sweeps one marketplace and returns cited facts only.
- Analyst agentsVerify one claim at its source, or write one section of a dossier.
- Verifier agentIndependently tries to disprove each claim before it counts.
- Reviewer agent and a personThe shortlist is decided by a strong model and approved by a human.
Our role: designed the agent roles, rules and review steps.
Web apps built and deployed
Small full-stack web apps built while mastering AI-assisted development: an interactive quiz app and a personal portfolio site, written in a modern React framework, tested with browser automation and deployed to the cloud.
- Modern React frameworkServer-rendered pages, fast and search-friendly.
- Browser-testedKey flows covered by automated browser tests.
- Cloud deploymentShipped to a global hosting platform with automatic deploys.
- AI-assisted developmentBuilt with AI pair-programming, reviewed by a person.
Our role: built, tested and deployed both.
What we work with.
Software & web
Web platforms, mobile apps, .NET, Java, PHP, JavaScript, enterprise CMS.
AI & agents
LLM workflows, multi-agent pipelines, prompt and evaluation design, human-in-the-loop approval.
Data & integrations
APIs, point-of-sale and accounting integrations, SQL data validation, payments.
Quality & compliance
Test strategy, regression, performance and load testing, release gates, regulatory rules.
Cloud & DevOps
CI pipelines, containers, DNS and email security, deployment and monitoring.
Delivery
Agile and Scrum, estimation, sprint and release management, stakeholder communication.
Have something to build, or finish?
Tell us about it. You'll get a clear plan and a fixed quote, and a senior engineer replies within one business day.