Healthtech · 720-hour cycle
kaedax builds HIPAA-aware healthtech software — clinician consoles, AI scribes, patient-facing workflows — in a fixed 30-day (720-hour) cycle. PHI handling, encryption, access controls, and evaluation harnesses for any AI output are part of the spec from day one, built by a team with petabyte-scale healthcare ML experience.
Scribe accuracy
94.2%
eval-gated
Time per note
−71%
vs manual
Active panels
1,842
+12% mom
PHI audit
100%
logged
[ §01 ] the problems we're hired for
PHI handling decided after the architecture is locked means re-architecture, not compliance. We design data boundaries, encryption, anonymization, and access controls into the first spec — our team has run HIPAA-compliant ML on a 2.5-petabyte clinical dataset.
Healthtech fails on workflow, not features. A console that adds forty seconds per patient encounter dies in pilot. We build with the clinician's loop as the unit of design and instrument it from day one.
An AI scribe or triage assistant that's wrong 6% of the time needs to know *which* 6%. Every AI surface we ship comes with an evaluation harness — graded test sets, regression gates, and accuracy budgets your clinical lead signs off on.
HL7, FHIR, and vendor-specific quirks routinely add quarters to roadmaps. We scope integration depth honestly on the scope call and build adapter layers that isolate vendor chaos from your product.
[ §02 ] what ships
The shapes of healthtech work that fit a 720-hour cycle. Every build ships with runbooks, monitoring, and 30–60 days of post-launch on-call.
Encounter workflows, patient timelines, order surfaces — designed around the clinical loop, not the org chart.
Ambient documentation, summarization, triage — every model output gated by an eval harness with clinician-approved accuracy budgets.
Onboarding, scheduling, remote monitoring surfaces with consent and PHI boundaries enforced in the schema.
Encryption at rest and in transit, ephemeral inference, anonymization pipelines, access audit logs.
[ §03 ] proof
Every agency we spoke to treated HIPAA as a change order. kaedax treated it as the spec. The eval harness on the scribe is the reason our clinicians trust it.
[ §04 ] questions
Yes, for a focused surface: one clinician console, one AI scribe, one patient workflow. HIPAA awareness is architectural — PHI boundaries, encryption, access controls — and we design it into the spec on day one rather than retrofitting it. Our CADUCEUS engagement shipped a clinician console with an AI scribe at 94% gated accuracy in one cycle.
PHI lives in your cloud, under your accounts, encrypted at rest and in transit. AI inference runs ephemeral — no patient data retained by model providers. Access is scoped per role and every read is audit-logged. We partner with your compliance counsel; we don't replace them.
Market quotes for HIPAA-compliant MVPs typically run $60,000–$250,000 over 3–6 months. kaedax prices per fixed 720-hour cycle, shared on the scope call. The fixed cycle removes the timeline-overrun risk that inflates most healthtech budgets.
Yes — with the evaluation harness built first. Every AI output is gated by graded test sets and regression checks that your clinical lead approves. We've written publicly about why the eval harness comes before the agent.
We scope EHR integration honestly: FHIR-first where the vendor supports it, adapter layers that isolate vendor quirks, and a clear call on the scope call about what fits in one cycle. Deep bidirectional EHR integration may need a second cycle — we tell you upfront, not at day 25.
A 15-minute scope call with a kaedax founder and the engineer who'd lead your build. We either fit your healthtech product into 720 hours, or we tell you why not — same call.
▷ other verticals