Engineering & Architecture
Working with AI coding systems is not straightforward. Designing systems that run with AI needs a specialised set of skills and understanding. We design, build and review AI systems that stay secure, maintainable and affordable to run as they grow.
Services
- Software architecture
- Platforms, services, data and integrations designed to be run, changed and understood by whoever looks after them in two years' time.
- Architectural assessment
- An independent look at how your system is built: where it will break, where it leaks, where it costs too much, and what to change first.
- Applied AI
- Putting LLMs and agents to work on real problems, with evaluation, guardrails and running costs worked out before launch, not after. On your own hardware, when the data can't leave the building.
Our flagship framework
Foundationist
A secure coding framework for software written with AI agents.
With the rise of AI, and seeing its inherent flaws, we decided to do something about it. We wrote Foundationist: a framework with defence in depth and secure code principles built in as standard.
We build using Foundationist, our innovative AI secure coding system. It allows rapid development, with ISO 27001 governance built in, and is designed to support NHS DSPT, DTAC and DCB0129 evidence and the needs of other regulated sectors.
For data protection, it flags where a DPIA is needed, maps where personal data goes, including to outside providers, and gives your data protection officer a live view of it, so GDPR reviews start from evidence rather than guesswork.
Code is built towards OWASP ASVS Level 3, the most demanding level of the Application Security Verification Standard, with progress tracked by an automated scorecard.
It has protection built in against 20 common attack types, covering all ten OWASP Top 10 categories, so users can't be tricked into actions they didn't intend, sent to fake sites or have their passwords guessed.
AI tools are good at a first version. The trouble starts with the next change: each session writes things a little differently, and security controls thin out or disappear. Foundationist takes those controls out of individual files and into the framework. Routes, APIs, permissions, secrets and data are declared in one place, and the build fails if anything isn't declared.
Ask about FoundationistQuestions we're often asked
What is Foundationist?
Foundationist is our secure coding framework for software written with AI agents. It builds in defence in depth and secure code principles as standard, so security doesn't slip as new features are added.
Can you run AI on our own hardware?
Yes. When data can't leave the building, we can run AI models on your own hardware, as we did for Caseo, which analyses legal evidence without sending any data to the cloud. See Research & Development.
What is an architectural assessment?
An independent look at how your system is built: where it will break, where it leaks, where it costs too much, and what to change first.
Do you only work on systems that use AI?
No. We design and review software architecture with or without AI in it, so systems can be run, changed and understood by whoever looks after them in two years' time.
Tell us what you're building
Describe the system, the deadline, and what's worrying you. We'll tell you what we'd look at first.
thearchitect@