Two siloed sites become one harmonized Domain Intelligence Layer. The maturity gain is measurable.
One patient, held in two systems that have never spoken: a Snowflake hospital lab warehouse and a Google Cloud physician-group EHR. At each site, the Semantic Classification Engine reads the raw values and the Data Mapper Model transforms them to standard FHIR R4 through a curated mapping profile, which carries the configured ontology set (ICD‑10, SNOMED CT, RxNorm, LOINC). The two standardized datasets then flow into the Domain Intelligence Layer, where the Data Harmonization Model identity-resolves and merges them into one longitudinal record, across sources and across time. All three engines run under Emergent Mapping Intelligence (EMI), the orchestration layer that decides what runs where and resolves disagreements between the models. The DERMS score below tracks exactly how much better the pipeline understands the data at every hop.
The pipeline
One patient, two silos, one record.
Live · starting pipeline
00Raw silos
01Classify · SCE
02Map via profile · DMM
→Standardized FHIR
03Harmonize · DHM
→The network
00 · Raw silo · Site A
Snowflake · Hospital lab warehouse
Site A · raw silo
Unlabeled tables. Local codes. No schema.
RIVERA, J · 03/12/1985 · MRN 30021
CREA-S 1.4 · mg/dL
intake: "diabetic"
ref: DR CHEN · NPI 1780043210
Hospital lab · silo 1 of 2
01 · Classify · SCE · Site A
AI model 01 · SCE
Semantic Classification Engine
Reads the unlabeled lab warehouse and binds each value to clinical meaning, with confidence.
Awaiting input…
Orchestrated by EMI
02 · Map · DMM · Site A
AI model 02 · DMM
Data Mapper Model
Maps to FHIR R4 through the site’s configured mapping profile.
Emergent Mapping IntelligenceEMI is the orchestration layer over SCE, DMM and DHM. It decides what runs where, resolves disagreements between the models, and scales the architecture across sites, so every mapping profile keeps improving in production.Evolutionary Neural Networks · USPTO App 19/181,522 allowed
The measurement
DERMS: Data Exposure & Reasoning Maturity Score
Accuracy tells you whether a pipeline succeeded; DERMS tells you why. It decomposes understanding into Epistemic Uncertainty (does the model know what it does not know, and it falls as classification and mapping add evidence), Data Familiarity (has it seen data like this, and it rises as profiles accumulate exposure), and Reasoning Competence (can it derive the correct transformation, and it rises as operator chains validate against golden truth). Watch it climb from the raw silo, to the standardized site outputs, to the harmonized network layer.