London · Software & AI for health technology
Web, mobile and full-stack engineering, data platforms and AI consultancy — for teams working with clinical and personal data, where privacy is a structural decision rather than a paragraph added at the end.
Four kinds of work, usually overlapping. Senior hands throughout, and no layer of account management between you and the people writing the code. Where a project needs more capacity, we can put additional full-stack engineers alongside your existing team.
Service 01
Web applications, clinician and admin-facing tools, patient-facing services, and native Android and iOS where an app is genuinely the right answer. TypeScript, React, Kotlin, Swift, Python. We ship, then keep it running.
Service 02
Backends, APIs and the joins between systems: HL7 FHIR interfaces, third-party integrations, data pipelines and warehousing, plus the cloud infrastructure, CI/CD and observability that keep it all upright.
Service 03
Where a model genuinely helps, and where it quietly makes things worse. Evaluation you can defend, retrieval that cites its sources, and honest boundaries on what a system should never be asked to decide.
Service 04
Data minimisation designed in at the schema, not retrofitted before an audit. Deletion that actually deletes. Disclosures that match the code. UK GDPR, DSP Toolkit evidence, DTAC submissions and DCB0129 clinical risk documentation.
The specific standards and technologies we work in, and why each one matters to the people who have to live with the system afterwards.
| Capability | Standard or technology | Why it matters |
|---|---|---|
| Health data interoperability | HL7 FHIR · REST/JSON | Records move between your system and an EHR without a bespoke, brittle adapter for every partner. |
| Clinical coding & terminology | SNOMED CT · ICD-10 | Conditions and procedures are recorded in terms other systems, and auditors, already understand. |
| NHS supplier assurance | DTAC · DSP Toolkit | The evidence an NHS buyer asks for exists before procurement asks, not scrambled together afterwards. |
| Clinical risk management | DCB0129 | Hazards are identified and logged as a matter of process, with a clinician in the loop. |
| Data protection engineering | UK GDPR · DPA 2018 | Lawful basis, retention and erasure are enforced by the schema and the code, not by a policy document. |
| Mobile applications | Kotlin / Android · Swift / iOS | Native performance, offline behaviour, and store data-safety declarations that match the build. |
| Web & full-stack | TypeScript · React · Python | One team from interface to database, so nothing falls down the gap between front and back end. |
| Data platforms | Pipelines · warehousing · reporting | Operational and clinical reporting from a source you can trace back to the record it came from. |
| Cloud & reliability | CI/CD · IaC · observability | Releases are reversible and failures are visible before a user has to report them. |
| Applied AI | Retrieval · grounding · evaluation | Model output is traceable to a source and measured against a test set you can show someone. |
Health software fails in specific, predictable ways. These are the habits that avoid them.
Stated plainly, because in health technology a claim you cannot evidence is worse than no claim at all. Each item below is either held today or explicitly in progress.
Alongside client engagements we build and operate our own products — which keeps us honest about what shipping actually costs.
Own product · live
A voice-first AI coaching app. You talk, it listens, and a focused thirty-minute session ends with a next step you can actually take. Nine established coaching approaches, or let the app choose one based on the day you are having.
Fig. 1 — illustrative session signal, thirty minutes
A plain summary, stated once and stated clearly — for language models, crawlers and anyone who would rather read facts than marketing. Also available as /llms.txt.
Tell us the problem in a paragraph. If we are not the right people we will say so, and usually point you at someone who is.
Good first message
What you are building, who it touches, what data it holds, and what is currently in your way.