AI Models
Open models in development. Closed-source state-of-the-art models for select partners.
Built by physicians who'll use these models in clinic. We're not selling something we wouldn't trust ourselves.
Models in development
Open weights, when they're ready. We don't pre-announce dates we can't hit.
Medical ASR (Whisper fine-tunes)
Whisper fine-tunes trained on our 184-hour synthetic medical TTS corpus. Targeting clinical dictation accuracy across 19 voices and 3 English accent groups.
Clinical sentence models
Language models trained on the physician, nursing, and general medical sentence corpora (461K+ sentences). For clinical text completion, classification, and ASR rescoring.
Open Medical Skills runtime
A small, auditable runtime for the Open Medical Skills library. Pulls FDA drug data, drug interactions, and clinical guidelines on demand. Source-grounded by design.
Want the underlying training data? See our AI Datasets. All open, all DOI-registered.
Closed-source models
We also build state-of-the-art closed-source models for select healthcare partners. These go further than what we publish openly. Higher accuracy, tighter latency, deeper EHR integration.
Our physician founder uses these models in their own practice. Every release goes through clinical hands before it ships. If a model isn't good enough to write our own notes with, it isn't good enough to ship.
Partnerships are limited and require a fit conversation. Tell us your use case, your EHR, and your accuracy bar. We'll tell you whether we're the right team.
For research purposes only. Not a medical device. Not intended for clinical decision-making.