Built for medical validation, not hype.
Kayalogy is engineered for transparent performance measurement, clinical evaluation, and regulatory-aligned development — ensuring outputs are reproducible, traceable, and audit-ready.
Important: Kayalogy is currently in development and not certified as a medical device. This page describes our validation approach and governance principles for research and evaluation use.
Clinical validation approach
Kayalogy is designed to support evaluation in controlled settings, with a focus on traceability, repeatability, and transparent reporting. We separate research outputs from any future certified clinical claims, and we document intended use and limitations throughout the development lifecycle.
- Non-clinical use (current status): research and evaluation only.
- Human oversight: outputs are designed to be reviewed by qualified professionals.
- Documentation: versioning, change control, and audit-friendly records as the platform matures.
Transparent evaluation pipeline
Models are trained and evaluated using reproducible pipelines. Evaluation outputs are versioned and traceable, enabling repeatable benchmarking and clear comparison across model versions.
- Standard metrics: Dice, IoU, Sensitivity, Specificity, Precision, F1
- Versioned runs, traceable artifacts, and documented configurations
- Controlled splits and consistent evaluation scripts (where applicable)
Intended use, limitations, and clinical context
Clinical validation requires clear boundaries. We define intended use and known limitations, and we aim to avoid overstating performance. Where feasible, we assess generalization, subgroup behavior, and failure modes.
- Defined intended use statements and scope (per project / model)
- Documented limitations and known failure modes
- Bias awareness and performance review across relevant populations (where feasible)
Risk management and quality discipline
We structure development around MedTech quality and safety principles. Risk analysis and change control are used to support safe iteration and a documentation pathway compatible with future regulatory expectations.
- ISO 14971-style risk thinking (hazards, mitigations, residual risk)
- ISO 13485-aligned process discipline as the platform matures
- Change control, traceability, and incident-aware improvement
Privacy-first data governance
Sensitive data requires strict governance. Kayalogy follows data minimization and access-control principles designed for Swiss FADP and applicable GDPR requirements, with security-focused handling and clear retention logic.
- De-identification expectations for clinical uploads
- Role-based access controls and audit-friendly security controls (where applicable)
- Retention and deletion principles aligned with purpose and legal constraints