Engage

Concrete ways to contribute, by role.

Guardian Angel is being built as patient-centered infrastructure. The work it requires is distributed across informatics, AI safety, policy, clinical research, and sustained funding. Below are the specific pathways for each.

Informatics

For the people who build the data layer.

Standards WG

Join the standards working group

FHIR profile refinement, SMART on FHIR scopes, multi-modal data extensions, consent receipts, audit-log schemas. The output is intended to be public and citable.

Reference impl.

Contribute to the reference implementation

The patient-controlled record, the ingest connectors, and the clinician-view SMART app are being developed as a reference implementation. Pull requests, design reviews, and security audits are welcome.

Eval

Build the evaluation harness

Synthetic and consented-cohort evaluation harnesses for trajectory detection, summarization fidelity, and recommendation grounding.

AI & safety

For the people who reason over the record.

Red team

Adversarial evaluation of agent outputs

Red-team the trajectory engine, the summarizer, and the recommendation grounding. Failure modes that should never reach a patient are the priority targets.

Grounding

Guideline-citation infrastructure

Build the infrastructure that ties recommendations to specific guideline statements and to specific record observations. Untraceable recommendations are an architectural defect.

On-device

On-device inference engineering

Apple Silicon and other on-device NPU work for low-latency conversational use, privacy-sensitive reasoning, and graceful degradation for low-resource deployments.

Policy

For the people who shape what is permissible.

Governance

Co-author the governance framework

Public-interest operating model, advisory composition, decision-making procedures, and patient-interest safeguards. Intended to be public and adoptable by other projects.

Liability

Liability and insurance position paper

The clinician-side and patient-side liability framework for AI-assisted care needs structured argument and engagement with insurers, clinical societies, and regulators.

Regulatory

Regulatory pathway engagement

Early engagement with FDA, EMA, Israeli MoH, and equivalents on intended use statements, SaMD classification, and AI Act risk tier alignment.

Clinical research

For the institutions that can run a cohort.

Cohort

Host a consented pilot cohort

Pediatric trajectory, oncology shared decision-making, complex chronic disease, and cross-border continuity are the highest-value initial targets. Integrated payer-provider systems are particularly well-positioned.

Evidence

Prespecified outcome measurement

Time-to-recognition for slowly evolving findings, clinician documentation burden, escalation appropriateness, patient-reported understanding and trust.

Multi-site

Multi-site recruitment infrastructure

The architecture's cross-institutional and cross-border recruitment capability needs to be demonstrated with real trials at real scale.

Funding

For the people who can underwrite public-interest infrastructure.

Guardian Angel's funding model is closer to public-interest infrastructure than to a conventional health-tech product. The current focus is on philanthropic, research, and integrated-system funding, with explicit governance to keep the patient interest primary as the project scales.

The highest-leverage uses of funding right now are: hardening the reference implementation, running the first consented pilot cohorts, building the evaluation harness, and developing the regulatory and governance position papers that the project will need to engage policymakers credibly.

Contact

Tell us what role you would play.

Submissions are reviewed by the project team. We will follow up within two weeks. No sensitive medical information should be included in this form.