Faithful to the source
Judged on omissions, unsupported statements and how much the physician has to fix.
Every model call in Lyra gets a structured record, one task and a person who signs off.
Four layers, designed together.
FHIR R4 connects each patient to their encounters, results and documents, so context has the same shape everywhere.
Lyra decides what context a task needs, what it should produce and where the result goes next.
Drafting, summarising and sorting are separate jobs, each with its own inputs and format.
Clinical content returns to the physician. Admin exceptions return to the responsible team.
Built for private Canadian-hosted infrastructure.
Tied to each user’s identity and team permissions.
Only the context a defined task needs.
Canadian residency required for data, prompts and output.
Clinical approval and accountable admin actions.
Lyra V1 is in testing and certification.
Read product statusWhat a model has to clear before it’s useful in a clinic.
Judged on omissions, unsupported statements and how much the physician has to fix.
Judged on time to a usable draft, capacity under load and cost per task.
Judged on data-access limits, approval points and exception handling.
Lyra’s architecture is designed for private Canadian-hosted infrastructure. Residency requirements cover stored records, prompts and model output.
Not yet. Lyra V1 is in testing and certification. We don’t claim completed certification or audited production controls. Read product status.
No. Lyra drafts and organises. A physician reviews, corrects and approves every clinical note before it is finalised.
Clinical data is structured as FHIR R4. Each connection uses authorised interfaces and validated data. With each clinic, we plan which systems to connect and which records to bring over.