Six things, and we say no to the rest.

Product engineering

End-to-end delivery of web platforms and portals, from the first architecture call to the thing running in production with users on it.

• Multi-tenant SaaS with real tenant isolation, not a shared database with a filter on it
• Client portals, admin tooling and the unglamorous CRUD that carries most of the value
• Design and build together, so the interface is not an afterthought bolted on at the end

• Next.js
• TypeScript
• Node
• PostgreSQL
• Supabase
• Prisma

Applied AI

AI as a feature inside a product people already use, built against evaluation harnesses and cost controls rather than shipped on vibes.

• Natural-language workflows over existing business processes
• Multi-provider model routing with fallback, so one provider outage is not an outage
• Per-tenant budget caps and spend tracking wired in from the start
• Evaluation harnesses, because "it looked right in testing" is not a release criterion

• LangGraph
• FastAPI
• OpenRouter
• Anthropic
• OpenAI
• fal.ai

Platform and cloud

The infrastructure and delivery pipeline underneath. Usually the part that decides whether the rest of it survives its first month.

• Azure Container Apps, Static Web Apps, Front Door, Blob and Table Storage
• Entra ID single sign-on and tenant-aware access control
• CI/CD, environment promotion and the runbooks that go with them

• Azure
• Container Apps
• Entra ID
• GitHub Actions
• Docker

Business systems and ERP

Getting an organisation off a stack of half-working tools and onto one system, including the migration nobody wants to own.

• Odoo implementation and configuration against real operational processes
• Consolidating point solutions for POS, quoting, specification and accounting
• Data migration, reconciliation and cutover planning

• Odoo 19
• Python
• PostgreSQL
• ETL

Microsoft 365, Copilot and Office engineering

Getting an organisation the capability it has already paid for in its Microsoft tenant, then building on top of it where the out-of-the-box answer runs out.

• Microsoft 365 Copilot rollout: tenant readiness, permission and data hygiene before switch-on, and adoption that survives the first month
• Copilot extensibility: declarative agents, plugins and connectors over your own content, and Copilot Studio agents where a conversation should replace a process
• Word, PowerPoint and Outlook add-ins on Office.js for the work Copilot will not do on its own
• Teams as an application surface rather than just a chat client: bots, message extensions and embedded tabs
• Power Platform where it genuinely fits, including Power Automate and Dataverse, and the judgement to stop and write real software when it does not
• SharePoint and OneDrive content architecture, Microsoft Graph integration, Entra ID identity and conditional access
• Modernising legacy VBA and VSTO estates without discarding the process knowledge two decades of them encode

• Microsoft 365 Copilot
• Copilot Studio
• Office.js
• Microsoft Graph
• Teams
• SharePoint
• Power Platform
• Entra ID
• Purview
• C#
• TypeScript

R&D and prototyping

Feasibility work, prototypes and pitch builds. Some of it becomes a product, some of it proves an idea does not work, which is also a result.

• Working prototypes against real constraints and real data shapes
• Technical due diligence and architecture options with the trade-offs named
• Internal tooling that earns its keep before we offer it to anyone else

• Whatever the question needs