Why we own our clinical ASR
Renting speech APIs caps accuracy and leaks PHI. Here's how we built Vocalis on Riva.
ReadEngineering · Clinical AI · Product · Company
Epic FHIRathenahealthNVIDIA InceptionHL7SOC 2HIPAA
Featured
Latest writing
Verifiers beat bigger models
In medicine, an independent verifier earns more trust than a larger generator.
ReadThe data flywheel, concretely
How every physician edit becomes labeled training data in Forma.
ReadCoding with the receipts
Evidence-linked ICD-10/CPT and why auditability is the feature.
ReadOn-prem inference for PHI
Packaging owned models as NIM microservices on NVIDIA AI Enterprise.
ReadDesigning motion that respects clinicians
Cinematic, accessible, never in the way of the work.
Read
Part 1
The wedge.
Why we finish the encounter instead of stopping at the note.
Part 2
The stack.
Owned speech, reasoning, vision — and why GPUs are essential.
Part 3
The flywheel.
How corrections compound into per-provider models.
Part 4
The deployment.
From cloud to air-gapped edge, without losing PHI.
Topics
Browse by topic
Find the writing relevant to your work.
Engineering
Models, serving, and infrastructure.
Clinical AI
Reliability, evals, and specialty depth.
Product
How the ecosystem fits together.
Security
PHI, on-prem, and governance.
Company
Team, mission, and milestones.
Research
What we're learning as we ship.
Cadence
The blog [PLACEHOLDER]
0
Pillar posts / month [PLACEHOLDER]
0
Topics
0
Build-in-public series
0
Specialties covered
Newsletter
Get it in your inbox
A short, technical newsletter on owned clinical models and agentic medicine — no fluff, roughly twice a month. Subscribe via the contact page; we mark anything unverified as illustrative.
Prefer RSS? A feed is available alongside the blog once published.
Posts2 / mo
Seriesbuild-in-public
Topics4
Toneno fluff
Why we write
Building in public builds trust
Healthcare buys on credibility. Writing openly about our models, evals, and failures is how we earn it — and how we recruit the engineers who want their work in exam rooms.
- Engineering depth, not press releases
- Honest about aspirational vs. shipped
- Clinical reliability in the open
- Recruiting through real work
Proof
What clinicians say
Outcomes are reported by design partners and marked illustrative [PLACEHOLDER].
“The clearest writing on clinical AI reliability I've found.”
“Build-in-public done with actual technical depth.”