Skip to content
How we work
Start a conversation Talk to us

For healthcare and life sciences

How we would work

Use health data and AI with privacy built in

A healthcare data platform starts from the sensitivity of the information: who may see it, what the purpose needs and where people must decide. We build data and AI for operational, administrative and research work. AI does not replace clinical judgment.

Free first conversation · A person replies within one business day · You own the code and accounts

Walk through privacy by design

  1. Classify the information first
  2. Collect and keep only what the purpose needs
  3. Control access by role and consent
  4. Keep clinicians in charge of clinical decisions
  5. Test, record and review

See the map →

01 How we work

How we would approach sensitive health data

  • Privacy as a design input

    Data is classified, minimized and pseudonymized where the purpose allows, before any pipeline is built. What is not needed is not collected, and retention follows your policy.

  • Purpose before platform

    Each use starts from a stated purpose and the people it serves, so the data collected, the access granted and the tests run all follow from it.

  • Evidence for your reviewers

    Access, consent where it applies, logging and testing are documented as we go, so your information governance and security teams review a written design from the start.

  • The frameworks you choose

    Designs can be aligned with ISO/IEC 27001, the NIST Cybersecurity Framework and your own policies. Alignment is not certification, and we do not claim any regulatory approval.

What we'll tell you

If the data is not ready, or a plain report would serve your staff better than a model, we will tell you before anything is built.

Controls map

Walk through privacy by design

Pick a step to see what we would do at that point and what your governance team would receive. The steps run in order, but you can start with any of them.

Which step would you like to see?

How it is handled

Classify the information first

We list the data the project would touch and classify it: identifiable health information, pseudonymized research data, operational records. The classification sets the controls for everything after it, so your information governance lead agrees it before any data moves.

  • A data inventory with a sensitivity class for each source
  • Controls chosen per class, not one setting for everything
Collect and keep only what the purpose needs

Each field has to earn its place. A capacity report needs counts by ward, not names. We minimize at the source, pseudonymize where the purpose allows and set retention to your policy, which also limits what a breach could expose.

  • Minimization at the source, before data is copied
  • Pseudonymization or aggregation where the purpose allows
  • Retention and deletion that follow your policy
Control access by role and consent

Access follows the roles in your identity provider, and consent is respected where it applies. An assistant over policies or records retrieves only what the person asking may already open, and every access is logged.

  • Role-based access checked before retrieval
  • Consent flags carried with the data where they apply
Keep clinicians in charge of clinical decisions

AI does not replace clinical judgment. We point it at administrative, operational and research work. Where an output could influence care, it is presented as support for a qualified person, with its sources and limits shown, never as a decision.

  • Clinical use only with the validation and authorization it requires
  • Sources and known limits shown with every output
Test, record and review

Before release we test against representative cases agreed with the people who will rely on the system, including the failure cases. After release, logs, monitoring and a regular review show whether it still behaves as agreed.

  • Test cases and thresholds agreed before release
  • Monitoring and a named owner after it

How it starts

  1. A first conversation

    You describe the data, the people who would use the system and the decisions around it. We ask about governance, consent and who signs off.

  2. A privacy and access design

    Classification, minimization, access and logging are written down and reviewed with your information governance and security teams.

  3. A bounded proof of concept

    One administrative or research use on approved data, with agreed test cases, thresholds and known limitations.

  4. Your decision on the results

    You get the results against the criteria and a plain recommendation, which may be to go no further.

The usual engagement model: Proof of concept or pilot

Talk this through with us →

Ready to talk it through?

Start a conversation

02 A typical situation

What this usually looks like

A hospital operations team might want one view of bed capacity and staffing, with the data spread across clinical and rostering systems and an information governance team rightly cautious about any new copy of patient data.

We would agree with governance which fields the view needs, which can be aggregated or pseudonymized, and who may see what. Then we would build pipelines with lineage and quality checks, and show any forecast clearly as a forecast, separate from what was measured.

03 Questions

Questions people ask

Not as a substitute for clinical judgment. We focus on administrative, operational and research work. Where an output could influence care, it supports a qualified person who stays responsible for the decision, and any clinical use needs the validation and authorization that apply to it.

We prepare for it from the first week. Your reviewers receive the data flows, access, retention and any subprocessors in writing before the build, and the work happens inside your environment where policy requires it.

Where your policy and the law that applies to you say. The platform runs in accounts or infrastructure you control, so location is a setting you own and can audit, not a promise from us.

Yes. Research data needs quality checks, provenance and lineage so results can be traced, plus pseudonymization or aggregation where the protocol requires it. We prepare and govern the data; research decisions stay with your team.

04 Start

Start with the data question, not the model

Book a call with an engineer. We will look at the data you hold, the governance around it and where AI could help staff without making clinical decisions.

What happens next

  1. You pick a time, or send us a message.
  2. A person at Nexstella reads what you shared before we talk.
  3. You leave with a clear next step, even when that step is to wait.

Or write to [email protected]. Send a brief instead

Meet

Pick a time with an engineer

What is your work email? The calendar invite and the video link go there.

The invite goes to the calendar you work from. Personal address? Use the Contact tab . If you stop before picking a time, we send one reminder to this address, and nothing else.

30 minutes, video call with an engineer · Times show in your time zone

What happens next

  1. You pick a time, or send us a message.
  2. A person at Nexstella reads what you shared before we talk.
  3. You leave with a clear next step, even when that step is to wait.

Contact

Send us a message

Pick the topic that fits and tell us what you are working on. A person replies within one business day.

A person replies from a real address.

Phone (optional)

Choose the country code, then type the number without it.

The topic leads the subject line, so the right person picks it up first.

A few sentences is plenty. You do not need to know which technology you need.

We use what you send to reply and, if it leads somewhere, to scope the work. Read the privacy notice, or write to [email protected].

Want the full picture? Read Healthcare and life sciences.

Also written for you: Financial services AI and data , Modern data platform .

Last updated