Software built by an engineer who supports it.
Mobile apps, websites, and web applications built with modern, proven, maintainable tech. Shipped on time, supported afterwards, and documented so any future developer can pick it up cold.
The stack we use
We choose the same handful of tools for almost every project, because we know them deeply and they cover 95% of cases. No "framework of the week" risk on your codebase.
Web
Django (Python) for back-end, server-rendered templates or React for richer front-ends, deployed via Docker. PostgreSQL or SQLite. Caddy or nginx in front.
Mobile
React Native (Expo) — one codebase, native iOS and Android. Pragmatic, well-supported, easy to hand to another team if you ever outgrow us.
APIs & data
Django REST Framework, OpenAPI specs, JWT or session auth. Background workers via Celery or RQ when needed.
Infra
Docker Compose on a managed VPS for most clients — fast, cheap, transparent. Cloud-native (AWS / Cloudflare) when scale demands it.
What we build
- Marketing websites with a CMS, contact forms and SEO baked in.
- Web applications — custom dashboards, internal tooling, customer portals.
- Mobile apps for iOS and Android, with proper App Store submission and CI.
- Integrations — Stripe, Mailchimp, HubSpot, anything with an API.
- Bespoke software for SMEs replacing a tangle of spreadsheets.
- AI features — LLM-powered search, summarisation, intake forms and assistants, built into your own product.
AI features in your software
More clients want LLM-powered features inside their own apps — document summarisation, smart intake forms, search-over-our-own-docs, drafting and triage. We build those properly: with cost, privacy and accuracy as design constraints, against the model that fits the job rather than whichever one is on a billboard this month.
- Claude, OpenAI or local models — we pick the API that fits your use case, your budget and your data-sensitivity profile, and we'll explain the trade-offs in plain English before you commit.
- RAG over your own data — your documents, your tone, your facts. The AI answers from what your business actually knows, not from generic training data.
- Cost-aware by design — prompt caching, smaller-model fallbacks, batched calls. We've shipped enough of these to know how to stop the bill running away from you.
- Privacy when it matters — for sensitive workloads we run open-weights models on a private VPS or on-prem. Your data doesn't end up in someone else's training set.
- AI-augmented build process — we use Claude in our own toolchain every day, which is part of why we ship faster than shops that don't.
Our process
Scope workshop
We map the must-haves, the nice-to-haves, and the things we'll deliberately leave out of v1.
Design & spike
Wireframes for the key screens, plus a technical spike on anything risky.
Build in sprints
Two-week iterations, demoed live. You see progress, you can change direction.
Launch & support
Deployed, monitored, documented. Optional care plan keeps it healthy.
Local engineer, on the doorstep.
Based in Marlow — we work across Marlow, Maidenhead, Henley-on-Thames, Beaconsfield, Bourne End, Cookham and High Wycombe, and the rest of the Thames Valley by arrangement. See where we work.