Axiomera
Enterprise Semantic Intelligence
The Transformation · Two Silos, One Layer

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
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.
profilesnowflake-lab-fhir
revisionsha256:9f3a71c2…
ontologyLOINC · SNOMED CT · FHIR R4
rolespatients · results · intake
Awaiting input…
Orchestrated by EMI
FHIR output · Site A
Standardized · Site A
FHIR R4 · validated · site-scoped
Patient · Julia Rivera
FHIR Patient · MRN 30021
Practitioner · A. Chen
FHIR Practitioner · NPI 1780043210
Observation · creatinine 1.4
LOINC 2160-0 · UCUM mg/dL
Condition · T2DM (intake)
SNOMED 44054006
FHIR dataset A → Domain Intelligence Layer
00 · Raw silo · Site B
Google Cloud · Physician group EHR
Site B · raw silo
Legacy codes. Free text. Different patient IDs.
JULIA RIVERA · F · 3/12/85 · BC-4411
250.00
metformin 500 BID
401.9
A1C 7.8
"SOB on exertion"
CHEN, ALICE MD · NPI 1780043210
Physician group · silo 2 of 2
01 · Classify · SCE · Site B
AI model 01 · SCE
Semantic Classification Engine
Reads legacy EHR extracts, ICD-9 codes, sigs and free text, and classifies every value.
Awaiting input…
Orchestrated by EMI
02 · Map · DMM · Site B
AI model 02 · DMM
Data Mapper Model
Same model, different mapping profile. The ontology configuration travels with the profile.
profilegcp-physician-fhir
revisionsha256:4c1e88d0…
ontologyICD-10 · SNOMED CT · RxNorm · LOINC · FHIR R4
rolespatients · problems · meds · labs · notes
Awaiting input…
Orchestrated by EMI
FHIR output · Site B
Standardized · Site B
FHIR R4 · validated · site-scoped
Patient · Julia Rivera
FHIR Patient · member BC-4411
Practitioner · A. Chen MD
FHIR Practitioner · NPI 1780043210
Condition · T2DM
ICD-10 E11.9 · SNOMED 44054006
MedicationRequest · metformin
RxNorm-coded · 500 mg BID
Condition · hypertension
ICD-10 I10 · SNOMED 38341003
Observation · HbA1c 7.8%
LOINC 4548-4 · UCUM %
Observation · dyspnea
SNOMED 60845006 · from free text
FHIR dataset B → Domain Intelligence Layer
03 · Harmonize · DHM
AI model 03 · DHM
Data Harmonization Model
Identity-resolves patients across sites, de-duplicates, rewrites references, unifies coding.
Awaiting standardized inputs…
0identities resolved
0duplicates collapsed
0clinical links made
2source sites merged
Orchestrated by EMI
The output · Domain Intelligence Layer
The output
Domain Intelligence Layer · harmonized across sources and time
One longitudinal record across lab + physician group. FHIR R4, provenance-tagged, governed.
Julia Rivera · one patient, two sites
tier-1 identity · 2 → 1 · references rewritten
Dr. Alice Chen · one practitioner
tier-1 NPI 1780043210 · 2 → 1 · lab + physician group
Type 2 diabetes · single problem entry
SNOMED 44054006 · lab requisition + physician dx merged
Creatinine 1.4 · linked to diabetes
LOINC 2160-0 · renal-function join
Metformin 500 mg BID
RxNorm-coded · linked to condition
Hypertension · evidence bound
SNOMED 38341003 · condition + vitals trail
HbA1c 7.8% · on timeline
LOINC 4548-4 · longitudinal series
Dyspnea on exertion · captured
SNOMED 60845006 · from free text
Build AI here ↓
Emergent Mapping Intelligence EMI 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.

DERMS = 0.30 · (1 − EU)  +  0.20 · DF  +  0.50 · RCtiers: auto-approve ≥ 0.85 · review ≥ 0.60 · tentative ≥ 0.50 · rejected below
Site A · raw siloS0
1−EU
DF
RC
no context · local codes only
Site A · standardized FHIR
awaiting
1−EU
DF
RC
converges per iteration
Site B · raw siloS0
1−EU
DF
RC
no context · legacy codes + free text
Site B · standardized FHIR
awaiting
1−EU
DF
RC
converges per iteration
Domain Intelligence Layer · harmonized
awaiting
1−EU
DF
RC
auto-approve 0.85
DERMS per harmonization commit · S2.1 → S2.final
cross-site identity + coding evidence compounds · uncertainty collapses
HeyDonto AI Technology · Axiomera · Codes and DERMS values illustrative