Startups
1-10 Employees
Rivertree HealthIrina is a health-tech startup developing a privacy-first, AI-powered solution for clinical documentation. Our goal is to reduce physicians’ administrative workload by supporting the creation of structured documentation while integrating securely into existing hospital workflows and IT infrastructure. We are currently developing and validating the solution together with healthcare professionals and hospital partners.
Details
Customer Problem
German hospital physicians spend around three hours per working day on documentation and administrative tasks. Clinical information is often captured retrospectively through fragmented hospital IT systems, dictation, copy-paste and repeated data entry; 52% of physicians report frequent redundant entries. This reduces time available for patients and contributes to incomplete discharge letters, information gaps during handovers and avoidable coding issues. Medical coding and controlling teams can only work with the information clinicians have documented, meaning missing clinical detail can also lead to billing disputes and lost revenue. Existing tools primarily transcribe text or review coding downstream, but do not ensure complete, structured documentation at the point of care.Business Model
Irina is sold to hospitals as an annual licence per inpatient department, starting at €30,000. The fee includes the on-premise appliance, software, models, offline updates, support and hardware replacement. The hospital pays a separate one-time fee for its KIS integration. There are no per-user fees or token charges. Once the appliance is installed, higher usage does not materially increase our inference costs.
Hospital sales fail when the first decision is too large. We therefore begin with a 90-day evaluation in one department. Irina and the hospital agree on the baseline and success metrics before the start. Standard evaluations are paid and credited against the first annual licence. Selected early reference sites may participate as research partners.
A successful evaluation converts into recurring revenue. One department represents €30,000 ARR, three departments €90,000, and a rollout across six departments €180,000 per hospital and year. The same IT and data-protection approval then supports expansion into other hospitals within the provider group.
Department-level analytics will be offered as an additional hospital-wide module. This creates a second recurring revenue stream from the same on-premise data layer. Integration is paid once; licences, support and expansion generate the recurring business.Technology Description
Irina is a sovereign clinical intelligence layer designed to operate entirely within hospital infrastructure. Rather than merely transcribing conversations, it captures clinical information once at the point of care and converts it into a longitudinal, source-traceable data layer that powers documentation, coding and operational intelligence throughout the inpatient journey.
Physician dictation, doctor-patient conversations, read-only hospital information system data and uploaded documents are processed locally through speech recognition, medical entity extraction and a model-agnostic orchestration layer for open-weight language models. A hybrid architecture separates generative from deterministic tasks: language models draft narrative sections, while medications, allergies, diagnoses, vital signs and codes are extracted from their sources and validated against official catalogues. Final documents are assembled programmatically.
Irina generates structured German drafts for admissions, ward rounds, handovers and discharge letters, alongside ICD-10-GM, OPS and aG-DRG coding suggestions. Every statement remains traceable to its underlying evidence; missing or conflicting information is flagged rather than invented, and physicians retain final approval.
The production architecture is a browser-accessible on-premise appliance with no patient-data egress and read-only integration through HL7v2, FHIR and ISiK. The resulting structured data layer also enables privacy-preserving department-level analytics, revealing documentation gaps, workflow bottlenecks, discharge delays and coding-quality patterns.
A functional MVP already demonstrates the core workflow, and the required hardware is available. The next stage focuses on completing local deployment, live hospital integration and prospective clinical validation.Market Description
Germany is the right first market for Irina. It combines scale with immediate economic pressure: 1,841 hospitals treated 17.5 million inpatient cases in 2024, and the German Hospital Institute reports that two out of three hospitals ended the year with a loss. Hospitals need to recover clinical capacity while protecting reimbursement.
We will initially target acute-care hospitals with 300–600 beds. Their case volume and DRG exposure are high enough for Irina’s impact to become visible in both working hours and euros. At the same time, most lack the resources to build and maintain their own clinical AI stack. Physicians use Irina, while department heads and medical controlling typically drive adoption. Hospital management or the provider group holds the budget.
Our pricing model starts at €30,000 per inpatient department and year. Applied to approximately 1,400 acute-care hospitals with six relevant departments each, this gives Irina a German addressable market of roughly €250 million in annual recurring revenue. The initial serviceable market is estimated at €150–175 million.
We do not need the entire hospital to commit on day one. A paid evaluation starts in one department, where we measure time savings, documentation quality and revenue effects. Successful results create a direct path into other departments and hospitals within the same provider group.
Reaching 40–80 hospitals within five years would generate approximately €6–14 million in ARR while leaving more than 94% of the German target market untouched. Mandatory ISiK interfaces create a standard route into hospital systems. Germany is the entry market; DACH and Europe follow through local adaptation of clinical language, coding catalogues and integrations.USP
Most AI scribes stop at a note. Irina builds a source-traceable clinical data layer for the entire hospital stay.
Coding tools work downstream, after the record has been written. They cannot recover clinical details that never entered the record. Irina captures them upstream through voice at the point of care, flags gaps while they can still be clarified, and reuses physician-approved information across admission, ward rounds, handover, discharge and coding. The same data layer supports department-level analytics.
The architecture limits what the language model is allowed to invent. Medications, allergies, diagnoses and other discrete facts are extracted from their source. Narrative sections use closed schemas, and missing information is marked as “not documented”. Each coding proposal is checked against the official BfArM and InEK catalogues and linked to its supporting evidence. A physician approves every output.
The production system runs inside the hospital on a department-level appliance and connects read-only to the hospital information system through HL7, FHIR and ISiK. One data layer connects documentation, coding and analytics without sending patient data outside the hospital or creating per-user cloud API costs.Finance Data
| Date | Investors | Amount | Stage |
Financial Overview
No revenue recorded yet