Built against SAHI — sutra by sutra.
India’s Strategy for Artificial Intelligence in Healthcare sets seven governing principles. Each maps to a property of the platform that can be tested, not to an intention.
Provenance is written in the same transaction as the finding it describes. If provenance cannot be written, the finding is not persisted.
Enforced, non-bypassable human sign-off. The model holds no authority; a named credentialed person does.
Full device obligations assumed regardless of classification, so development proceeds without regulatory gambling.
Subgroup performance, calibration and refusal rate reported in every validation. Refusal rate by subgroup catches a safety feature becoming an access barrier.
The signature binds a named, credentialed person to a content hash of exactly what was signed. Any change invalidates it.
The criteria engine emits every criterion met and unmet, with the data that decided each. Two interactions to evidence.
Fail-closed behaviour, drift monitoring, and a predetermined change control plan with automatic rollback.
Ready for independent benchmarking.
BODH — the Benchmarking Open Data Platform for Health AI, launched alongside SAHI at the India AI Impact Summit 2026 and developed by IIT Kanpur with the National Health Authority — is a federated environment under ABDM. Developers train models on-site without accessing raw patient data; only refined model weights return. It assesses performance, robustness, bias and generalisability against diverse, anonymised Indian datasets.
That is a description of the architecture we already committed to. A platform that centralises data cannot enter that process; one that is federated, standards-native and reproducible by construction can enter it immediately.
IIT Kanpur has described BODH as having the potential to become a default third-party algorithmic auditing tool for health AI. Soma Nova intends to submit for benchmarking rather than avoid it.
Aligned to the national rails.
Discuss national deployment.
For programmes, ministries and public-health bodies working within ABDM.