We bring the capability to your data, under your own controls.
Where your data lives, which models run, who can reach what, and how we answer your security and compliance reviews.
We work with the systems you already have.
We work across your core systems and the platforms you’ve already bought, and replace none of them.
We start read-only, through each system’s supported interface – its API, a reporting copy or a scheduled export – with accounts your IT team creates and can revoke. Nothing is installed on your production systems, and nothing reaches your operational networks unless your architecture allows it.
Production capability runs where your regulated data already sits, at the edge, on site or in your cloud, under your own identity, network and change controls.
- To start
- Read-only access
- How we connect
- Supported interfaces, with accounts your IT team issues and can revoke
- Write access
- A separate decision, with its own testing and approval
- Where it runs
- In your environment: edge, site or cloud, as the data allows
- Sign-in
- Through your identity provider, commonly Microsoft Entra ID
Where our own environments run
We hold as little of your data as we can, on purpose.
- Model
- Client-deployed. Capability is deployed into your environment rather than hosted centrally.
- Our tenancies
- Development, training and evaluation run in project environments on Microsoft Azure and Google Cloud Platform, and on Amazon Web Services where a project needs it.
- Hosting countries
- Ireland, Belgium and the Netherlands.
- Per project
- Compute, storage, repositories, secrets and service accounts are provisioned for each project, with least-privilege access approved by its owner and reviewed at defined gates.
- Code
- A private, organisation-wide repository with branch and review controls, and dependency and vulnerability checks.
Your data is used for one purpose, and kept apart.
We agree the architecture, hosting location, retention and access for each engagement.
- Nothing needed for a demonstrator
- You can see a demonstrator without sharing any data. A prototype that uses your data waits for a recorded data-protection decision.
- Kept apart
- Each project has its own segregated environment and named identities, with no shared data and no access across clients.
- Used for one purpose
- Your data is processed only for the contracted purpose, with its lawful basis, location, retention and sharing limits set out in the data plan and the contract.
- Never reused elsewhere
- Nothing is reused for anyone else without a new legal basis and your explicit authorisation. What we carry between clients is method, never one client’s judgement.
- Encrypted and logged
- Encrypted in transit and at rest, with secrets in approved secret-management services, never in code, and logs kept for investigation.
- Removed at project close
- When a project closes or someone’s role changes, access is removed and resources are transferred, archived or deleted as the contract sets out.
We use frontier or local models, depending on the task and your data.
We’re model-agnostic by design, so nothing we build depends on a single provider.
- Approved before go-live
- For each system we name the model, the service it runs through, the region and how long prompts and outputs are kept, and you approve them before go-live.
- Never a silent swap
- We assess a replacement model for its intended use, get it approved and re-run the reference tests before it goes live.
- The model that tests best
- At Meissner a purpose-built statistical model beat the large language models we tested, so that’s what runs.
- When the data can’t leave
- Where data sovereignty rules out a frontier provider, we run capable models on local hardware.
Are the models training on our data?
Not if the contract and set-up say no. We confirm both for each model before it sees your data, and again whenever the model or provider changes.
What will the tokens cost?
We match the model to the job and measure the cost per workflow. Run cost is part of the business case, and we report it once the system is live.
Security and compliance you can check.
We assess security, privacy and data classification before anyone gets access or anything is released. Send us your security questionnaire and we’ll answer it in full, with the policies your supplier assessment needs.
- Least privilege for named people
- Named identities and role-based groups, approved by an accountable owner. Privileged access is restricted and reviewed, with multi-factor authentication wherever it’s supported.
- A documented response
- Incidents follow documented escalation, containment, notification and lessons-learned procedures.
- Policies with owners and review dates
- Information security, data protection, AI governance, access control, incident management, business continuity, supplier management and more, each with an owner, a version and a review date.
- GDPR and a Data Protection Officer
- A Data Protection Officer, a register for data management, and procedures for personal and sensitive data, under our Data Protection Policy.
Our information security is aligned with ISO/IEC 27001. For biopharma, we build systems to support your GMP Annex 11 and 21 CFR Part 11 obligations; for medtech, your ISO 13485 quality system and 21 CFR Part 11. Both follow GAMP 5 practice. Your quality team decides how each system is classified and validated, and changes go through your change control as well as ours.
After go-live, we look for wrong answers as well as failures.
Support fixes failures. Wrong answers show up in overrides, deferrals and evaluation scores, and go to review. The team that built the system supports it under contract, with service levels, and we agree system by system who looks after it: us, your people or both.
Each system is designed around your operation.
We build it with your people, on your systems and data, around the decisions your operation makes.
- Built for your team
Designed for your own team to understand and extend, and handed over with its requirements, design, test evidence and traceability.
- Open records
CHAP is an open protocol, published as a public draft, so the records it keeps stay readable if you change vendors.