AI-native Clinical Research OS for research centers

Protocols should compile into operations.

LIND47 turns a trial protocol and its manuals into your site's daily work, with every action traced to its source page.

Request early accessSee how it works

Pre-launch. Built by a principal investigator. Seeking the first research sites.

Protocol → cited factsExample

6.1 SCHEDULE OF ACTIVITIES Screening occurs within 28 days before Cycle 1 Day 1. Study drug is given on Day 1 of each 21-day cycle. A complete blood count is collected before each dose.

Screening window: Day −28 to −1p. 41
Dosing: Day 1, every 21 daysp. 41
Lab: CBC before each dosep. 42
Reviewed and signed by the study team
The problem

Sponsor systems are built for oversight. LIND47 is built for execution.

The trial is designed elsewhere and executed at the site. LIND47 is one system for all your studies, whoever the sponsor. It does not replace the sponsor's EDC or IRT.

76%

of protocols are amended at least once, and the site has to absorb every change. Getz et al., Ther Innov Regul Sci, 2024

Every day, a site has to answer
  • Which protocol version is active?
  • Which visit window is closing?
  • Which kit expires soon?
  • Who is trained and delegated for this procedure?
  • What did the last amendment change?
  • What proves we ran the study correctly?
Founder

Built by an investigator who runs the trials.

Abel Costa, founder of LIND47
Abel Costa, MDFounder, LIND47
“I have run trials where the protocol lived in a PDF and the operation lived in spreadsheets, vendor portals and memory. Every year brings more documents and more systems to fill in. LIND47 is the one system I wanted for my whole team.”
Practice
Hematologist, São Paulo, Brazil
Research
Principal investigator in phase 1 to 3 oncohematology trials: lymphoma, CLL, multiple myeloma
Product
Designs and builds the LIND47 platform himself
The name
James Lind ran a fair test of scurvy treatments in 1747
How it works

One study, from the first PDF to the next amendment.

Six stages, illustrated with synthetic data.

01 · INGEST

Every study document, versioned.

The protocol, manuals, consent forms and amendments live in one registry, so the site knows which version is in force.

  • Protocol and amendments
  • Pharmacy manual
  • Laboratory manual
  • Consent forms
Study documentsIllustration
Protocolv3.0 · 142 pp · parsed
Pharmacy Manualv2.1 · 38 pp · parsed
Laboratory Manualv1.4 · 56 pp · parsed
Informed Consent Formv2.0 · 14 pp · parsed
Protocol Amendment 2received · impact pending
02 · COMPILE

Documents become cited facts.

Visits, procedures and eligibility criteria are extracted as structured facts, each pointing to its page. Contradictions are flagged, not guessed.

  • Schedule of activities
  • Eligibility logic
  • Gap map
  • Conflicts flagged
Schedule of activitiesIllustration
ProcedureSCRC1D1C1D8C2D1EOT
Informed consentX
Physical examXXXX
Complete blood countXXXXX
Study drugXX
PK sampleXX
14 activities, each citedpp. 41–42
Conflict: PK timing, §6.1 vs §8.3needs review
03 · REVIEW AND PUBLISH

People decide. Then they sign.

The study team reviews facts beside their source. The package is published only after approval and electronic signature.

  • Review workbench
  • Corrections tracked
  • Electronic signature
  • Versioned study package
ReviewIllustration
Screening window: Day −28 to −1fact

"Screening occurs within 28 days before Cycle 1 Day 1."Protocol v3.0 · §6.1 · p. 41

Confirmed by coordinatorreviewed
Approved by Principal Investigatorapproved
Signed electronically · pi.demo · 2026-03-12 14:02 (UTC−3)
Role: Principal Investigator · Meaning: approval of study package v1.0
Study package v1.0 published
04 · OPERATE

The package becomes the day's work.

Visit windows, tasks, pharmacy and lab queues come from the published package. If a rule is not met, the task is held for a person to resolve.

  • Agenda and visit windows
  • Enrollment and consent
  • Pharmacy
  • Lab and specimens
  • Training and delegation
Today · CoordinatorIllustration
  • Participant 0007 · Cycle 2 Day 1Visit window Day 20–24
    Closes tomorrow
  • Lab · CBC before doseResult required to release dosing
    Ready
  • Pharmacy · prepare study drugSecond verification required
    Queued
  • Participant 0012 · ScreeningConsent form v2.0 required
    Ready
  • Participant 0003 · PK samplingAssigned staff not delegated for PK
    Reassign
05 · PROVE

Evidence that builds as the study runs.

Each event is recorded with the package version and the source page behind it.

  • Audit trail
  • Deviations and CAPA
  • Site file
  • Monitor view
Audit trailIllustration
  • lab.resulted · Participant 0007 · CBC before dose
  • drug.dispensed · lot L-2204 · 2nd verification
  • visit.completed · Participant 0007 · C2D1
  • └ per package v1.0 · source: Protocol v3.0 §6.1 p. 41
  • deviation.opened · PK sample drawn outside window
  • └ CAPA assigned · quality manager
Each event traceable to its source page
06 · AMEND

When the protocol changes, see what changes.

A new amendment is compared with the package in force. The site sees which facts changed and what follows for consent, training and visits.

  • Amendment impact
  • Re-consent
  • Retraining
  • New package version
Amendment 2 · impactIllustration
Screening window: 28 → 21 days§6.1 · p. 41
PK sample added at Cycle 2 Day 1§8.3 · p. 58
Re-consent required4 participants
Training to repeat3 staff
Visits to reschedule2 visits
Held until ethics approval and PI signature
Who uses it

One signed package. Six roles.

What the package drives, role by role.

  • Coordinator

    Today's agenda, visit windows, due tasks, pharmacy and lab queues.

  • Principal Investigator

    Enrollment, deviations and site health at a glance.

  • Pharmacist

    Preparation, dispensing, accountability and temperature log.

  • Laboratory

    Kits, collections, specimen custody and shipping.

  • Quality

    Site file completeness, deviations and CAPA.

  • Monitor

    A read-only remote view of coded records.

AI-native, inside regulated boundaries

AI drafts. Your team decides.

  • No fact without a source

    Each one links to a document, version and page.

  • No operation from a draft

    Only signed packages reach daily work.

  • No agent writes regulated data

    AI proposes. People review, approve and sign.

  • When uncertain, it stops

    It fails closed and asks a person.

Design references, not certifications: ICH E6(R3), 21 CFR Part 11, LGPD.

Where AI is allowed to work

Protocol intelligence

Protocols, manuals and cited facts. No participant data.

AI works here

Site operations

Packages, visits, tasks, inventory.

Only through governed tools

Participant data

Participant records, in a separate database.

Closed to external AI models by design
Where we are

Built and tested internally. Not yet in use at a site.

Not yet commercially available, and not yet independently validated.

Working today

Documents to reviewed draft

  • Protocol and manuals to cited facts
  • Human review beside the source
  • Running on synthetic data
Internally tested

The operating modules

  • Acceptance test passed in pre-production, on synthetic data
  • Electronic signature and access control exercised across all user roles
  • Agenda, consent, pharmacy, lab, quality, monitor view
Ahead

First site evaluations

  • Study preparation with the first evaluation sites
  • Qualification for use with participant data
  • EHR and EDC connections
Early access

Evaluate LIND47 with one of your studies.

We are looking for a few research centers for the first evaluations. Tell us about your center and we will scope the evaluation together.

Email the founder abel.costa@lind47.com

You will hear back from the founder.