ORBE Health
The AI platform for healthcare information.
Two platforms built for the realities of healthcare in Colombia. One structures the clinical information institutions already have; the other captures what territories without connectivity still lack.
Auditable by design · Ley 1581 · Built for Colombia
Capabilities describe the intended scope of each final product — current status is noted on each platform.
MedSigt
In pilotTurns an institution's clinical documents into a validated, auditable database — and generates the regulatory reports from it.
- Ingestion with full provenance. Digital and scanned records, duplicate detection, automatic classification — every document traced to its origin.
- AI extraction, privacy first. Personal data is redacted locally before any text reaches a model; re-identification never leaves the institution.
- Normalization to standards. ICD-10, SNOMED CT, CUPS and national registries, so information is comparable across periods and specialties.
- Explainable validation, human review. Deterministic rules flag what is doubtful; reviewers see each field next to the highlighted evidence in the source document.
- Reports as projections of the database. The F207 report today; RIPS, MIPRES and FHIR interoperability on the same foundation next.
- Diagnosis
- Essential hypertension
- ICD-10
- I10 ✓ valid
- Insurer
- Nueva EPS ✓ normalized
- Age
- 15 months → early childhood
- Entry point
- → human review
Klassian
In pilotOffline-first health data collection for territories without connectivity — and the consolidation work that today happens by hand.
- Works with zero signal. Collection runs on low-cost Android devices; data is stored on device and syncs itself when connectivity returns.
- Ready-made instruments. The family characterization form (Resolución 3280) digitized with skip logic and validation — pick the form and start.
- Longitudinal household registry. What changed in a family between one visit and the next, household by household.
- Consolidation without spreadsheets. A unified live database: duplicate queue, per-collector quality, coverage by vereda.
- Exports in the required format. One click produces the file institutions must report upstream.
Tablet 04 · 12 forms queued · no signal — syncs when connectivity returns.
What ORBE is building
One foundation sits under both platforms: speech, extraction, coding, and generation, built for Spanish-language clinical work and for institutions that have to answer to a regulator. We build that layer so an institution gets a working application instead of a research project.
Speech to Text API
Medical speech recognition built for Spanish clinical conversation, available over an API and fast enough to run during the visit.
Text Generation API
Turn transcripts, findings, or structured data into the clinical documentation an institution actually has to produce.
Document extraction
Clinical PDFs and scans become structured fields, each one traced back to the evidence it came from.
Coding and normalization
ICD-10, SNOMED CT, and CUPS applied consistently, so records stay comparable across periods and specialties.
Offline data capture
Collection that runs with no signal on low-cost Android devices and reconciles itself when connectivity returns.
Regulatory reporting
Every report is a projection of the validated database, so the same numbers reconcile across F207, RIPS, and MIPRES.
Backed by research
The platforms come out of a research line we work on in the open — medical imaging, uncertainty, multimodal models, and clinical agents.
Learn moreSafety at ORBE
Built against the Colombian framework: Ley 1581 for personal data, Resolución 1995 for clinical-record confidentiality, Resolución 866 for auditability, and Resolución 3280 for Klassian's household form.
Join us
We are looking for institutions, clinicians, and researchers to build this with. If that sounds like you, write to us.
Get in touch