Build trust through evidence, not product theatre
RetGuard is a Delaware-based, research-stage retinal AI program. Our current work is a suite of separate classifiers studied on public fundus and OCT datasets and described in a Version 1 manuscript submitted to medRxiv. Public DOI assignment is pending screening.
We do not manufacture a retinal camera, do not claim an imaging manufacturer partnership, and do not offer a clinically cleared screening product. We are building the partnerships and evidence required to determine whether this research can translate safely.
Our operating principles
- Report failed and incomplete evaluations alongside positive results.
- Separate retrospective model performance from clinical readiness.
- Validate on a partner's specific device and population before product claims.
- Treat regulatory, workflow, human-factors, and quality evidence as required work, not footnotes.
A concrete starting point for collaboration
A partner is not being asked to buy a finished device. The first engagement is a scoped validation project built around available research artifacts and the partner's imaging domain.
RetGuard contributes
- ✓Research model weights and inference source
- ✓Model cards, checksums, and manuscript evidence
- ✓A predefined claim and evaluation boundary
- ✓Technical collaboration on preprocessing and analysis
A validation partner contributes
- ✓Target-device images and acquisition expertise
- ✓Independent clinical reference standards
- ✓Population and workflow context
- ✓Prospective, regulatory, and product-development capability
The next milestone is independent replication
We are prioritizing collaborations with ophthalmology groups, retinal imaging manufacturers, OCT manufacturers, and biostatisticians who value transparent validation.
Discuss a partnership