Now launching

Meet Sherpy.

Why “Sherpy”? →
sherpy Book a sprint

For digital health & medtech · Switzerland

Prove your AI uses health data lawfully.

Now that AI agents act on sensitive data everywhere, compliance has to be auditable — not assumed. Sherpy builds consent into the data itself, so agents can safely use more of it, and every access becomes evidence you can show auditors, partners and investors.

Evidence trail

wearable-app · production

Illustrative example
  1. Consent granted

    09:14

    heart-rate · purpose: model training

  2. Residency check

    09:15

    stored in CH · transfer to EU allowed

  3. AI agent access

    11:02

    triage-assistant · read · logged

  4. Consent revoked

    16:40

    agent access blocked instantly

Export for auditor

evidence-pack.pdf

Built on Pryv.io

Open-source personal data infrastructure, recognised as a Digital Public Good.

15+ years in health data

Consent, personal data lifecycle and regulated systems — hands-on.

Swiss by design

Data residency and sovereignty treated as engineering, not paperwork.

01 · Why now

AI is moving faster than the rules. The rules are catching up.

Agents

AI agents now touch personal data

Assistants connect to records, devices and CRMs. Few teams can prove who consented to which agent seeing what — and for what purpose.

Regulation

The rules are becoming enforceable

EU AI Act, MDR, the Swiss FADP, ISO 42001 and the European Health Data Space are moving from principle to audit.

Proof

Checklists are free. Proof is not.

Any team can generate a GDPR checklist. What earns trust is evidence produced by the running system — and signed off by people.

The hidden cost

Unused data isn’t neutral. It’s Data Debt.

Years of digital growth leave outdated, siloed data that nobody can use — or prove they may use. It still costs you, every day — and AI agents can’t touch what nobody can vouch for.

Data Debt is a trust problem. Here’s the way out
  • Rising IT & storage costs

    You keep paying to store and secure data that creates no value.

  • Slower decisions & growth

    Siloed, unclear data blocks the AI projects that should use it.

  • Compliance exposure

    Data with no clear consent or purpose is a liability waiting for an audit.

  • A bigger carbon footprint

    Unused data still consumes energy, every day it is stored.

02 · Our approach

Consent isn’t the brake on AI. It’s what makes data usable.

The most valuable health data sits unused — not because it lacks value, but because no one can prove it may be used. When consent travels with the data, AI agents can work with far more of it, safely and provably.

  1. 01

    Person

    Grants consent

    For a purpose, a scope and a duration — and can change their mind at any time.

  2. 02

    Data

    Carries its permissions

    On Pryv.io, every record knows who may use it, for what, and until when.

  3. 03

    Agent

    Asks before it reads

    An AI agent gets exactly what the purpose allows. Withdrawal stops access instantly.

  4. 04

    Value

    Data becomes usable

    More data can safely power AI — with the proof attached, not reconstructed later.

Data available to AI

Only what legal has cleared, case by case

Everything people agreed to, for that purpose

A new AI use case

New project, new legal review, months lost

Ask people, record consent, start using it

Someone withdraws

Data lingers in copies and pipelines

Agent access stops instantly — and it’s logged

Proof for an audit

Reconstructed from emails and spreadsheets

Produced continuously, as the system runs

Data Debt

Grows every year — stored, paid for, unusable

Shrinks — data is put to work, or deliberately deleted

03 · Use cases

Where the route gets steep

Typical situations we help teams navigate — from the first design decision to the moment an auditor, partner or investor asks for proof.

Prevention

A prevention app combining wearables and lab results

Challenge
Sleep, activity and biomarker data are merged to give AI-driven health insights. Users consented to “coaching”, not to model training.
Route
Map every signal and inference, split consent by purpose, add residency rules for lab partners.

Evidence you get

  • Consent ledger per user, signal and purpose — every grant and withdrawal timestamped
  • Inference register: which signal combinations reveal health data, and how each is protected
  • Data-flow map covering every lab and wearable partner
Answers GDPR Art. 7(1), 9, 30 · Swiss FADP Convinces Investors & partners in due diligence

Medtech · SaMD

An AI diagnostic tool heading for MDR and the AI Act

Challenge
A small team must show how training and patient data are governed, before a notified body asks.
Route
Readiness Sprint against MDR, the AI Act and ISO 13485; data lineage built into the pipeline.

Evidence you get

  • Training-data lineage: source, consent basis and version for every dataset
  • Data-governance chapter ready for the technical documentation
  • Gap report per framework, with owners and deadlines
Answers EU AI Act Art. 10–11 · MDR Annex II · ISO 13485 Convinces Notified bodies

Insurance & wellbeing

A wellness programme using behavioural data

Challenge
Members share step counts and habits for rewards. The data must never silently flow into pricing.
Route
Purpose-bound consent per signal, technical separation between rewards and underwriting.

Evidence you get

  • Tested separation: behavioural data has no technical path to underwriting
  • Access log of every read of member data, by system and purpose
  • Member-facing view of what is shared, with whom, and why
Answers GDPR Art. 5(1)(b), 9 · Swiss FADP Convinces Regulators, auditors & members

AI agents

A triage assistant reading patient records

Challenge
An AI agent needs access to records across systems — and patients may withdraw at any time.
Route
Agent access routed through Pryv-managed consent, scoped per purpose and revocable instantly.

Evidence you get

  • Every agent access logged: which record, when, under which consent
  • Consent withdrawal enforced instantly — and visible in the log
  • Agent permission scopes reviewable per purpose
Answers EU AI Act Art. 12 · GDPR Art. 7(3) Convinces Hospital DPOs & clinical safety

Lead, don’t follow. The first teams to prove trusted AI set the bar their competitors will be measured against.

Book a sprint

04 · Where we work

Health data — and the context around it.

Modern health products run on context, not just diagnoses. Sleep, movement, nutrition, mood and daily routines can reveal as much about a person as a medical file — yet they are rarely governed like one. Combined by AI, they become health data in all but name.

SleepActivity & movementNutritionMood & stressDaily routines & locationAt-home tests & biomarkersLifestyle & purchases

Who it’s for

  • Digital health apps
  • Prevention & wellbeing apps
  • Medical device software (SaMD)
  • Wearables & remote monitoring
  • Health insurers’ digital services
  • Clinical & research platforms

Frameworks

GDPRSwiss FADPEU AI ActMDR / IVDRISO 13485ISO 27001ISO 42001HIPAA

05 · Offer

Start with a sprint. Stay with proof.

How every engagement runs

  1. 01

    Map

    Trace how sensitive data is collected, stored, moved and used — including by AI models and agents.

  2. 02

    Bound

    Define consent, purpose, residency and regulatory boundaries for each data flow and jurisdiction.

  3. 03

    Build in

    Turn those boundaries into system-level controls, so governance is inherent rather than applied.

  4. 04

    Prove

    Generate continuous, verifiable evidence for auditors, notified bodies, partners and investors.

Start here

Readiness Sprint

2–4 weeks · fixed price

We map your data flows and AI use against GDPR, the Swiss FADP, the EU AI Act and MDR, show where Data Debt is building up, and hand you a prioritised roadmap.

  • Data-flow, consent & Data Debt map
  • Gap report per framework
  • Expert-reviewed, AI-accelerated

Then

Integrity Layer

Subscription

Continuous checks connected to your system, producing an always-current evidence pack instead of a yearly audit scramble.

  • Continuous readiness checks
  • Audit-ready evidence exports
  • Alerts when a flow drifts

In development · 2027

Consent Gateway for AI agents

Early partners wanted

AI agents reach personal data only through Pryv-managed consent — every access scoped, logged and revocable.

  • Dynamic, revocable consent
  • Per-purpose scoped access
  • Complete access audit trail

Sherpy assesses readiness and produces evidence. It does not replace legal advice or certification by a notified body.

06 · About

We’ve solved this before. Inspired by Pryv.io.

With a remarkable team, we turned consent into infrastructure. Now AI agents need the same answer: who may use this data, and for what? We’re not testing an idea — we’re scaling one that works.

Mission

Ensure transparent, ethical and verifiable use of sensitive data in AI systems — so organisations can innovate with it, and people can trust it.

Vision

A world where using personal data in AI is as trustworthy as online banking: understandable, accountable, and under the control of the people it describes.

Why “Sherpy”?

A sherpa doesn’t climb the mountain for you. They know the route, carry the load and bring everyone home safely.

That’s our role between your AI ambitions and the rules that protect sensitive data. You lead the climb — we make sure every step can be shown.

The team

AI-accelerated. Human intelligence at the core.

Models can draft a checklist in seconds. Knowing which data flows really matter, what a regulator will ask, and where people’s trust is at stake takes subject-matter experience. That judgement is what we bring — and sign our names to.

Portrait of Pierre-Mikael Legris

Pierre-Mikael Legris

Co-founder · Architecture & technology

Co-founder of Pryv SA and architect of Pryv.io — now open source and a UN-endorsed Digital Public Good. Data governance at Swisscom Digital Technology; today CTO of the Health Data Safe (HDS) Foundation and a patients’ advocate. 15+ years designing personal and health data systems where consent, access and auditability are built in from the first line of code.

Data & consent architectureHealth data infrastructurePatients’ advocacy
Portrait of Evelina Georgieva

Evelina Georgieva

Co-founder · Growth, partnerships & governance

Co-founder of Pryv SA, whose Pryv.io is now open source and a UN-endorsed Digital Public Good. Co-created the Data Governance practice at Swisscom Digital Technology. 15 years and 50+ projects turning regulation — GDPR, the Swiss FADP, HIPAA — into business models across digital health, pharma, insurance and the public sector.

Data governance & privacyGo-to-market & alliancesDigital health & life sciences

Open foundations

Trust you can inspect, built on Pryv.io

Pryv.io is Swiss, open-source infrastructure for personal and health data. It is fully open source, maintained by the non-profit Pryv Association, recognised as a Digital Public Good and supported by sponsors including the Health Data Safe Foundation.

Sherpy builds on Pryv.io rather than owning it. Your trust layer rests on open, inspectable foundations, and the public good stays public.

About Pryv.io Read the white paper

Consent built in

Dynamic, granular consent that can be granted, scoped and revoked at any time.

Access you can audit

Every read and write is attributable to an app, a person or an AI agent.

Open source, non-profit

Maintained by the non-profit Pryv Association and recognised as a Digital Public Good — no black box.

Deploy where data must live

Self-hostable, so data can stay in Switzerland, the EU or your own infrastructure.

Next step

AI won’t be limited by intelligence. It will be limited by trust in data.

Tell us what you’re building. In a first call we’ll show you where your data flows stand, where Data Debt is holding you back, and what a Readiness Sprint would cover.