Cohort intelligence
GPU-accelerated patient cohort identification, grouping, and comparison at health-system scale.
Product · Population health intelligence
Nexar —Encounter data at population scale.
Nexar transforms Scribara's growing encounter corpus into RAPIDS-accelerated population analytics and NeMo-trained predictive models — surfacing which patients are deteriorating, which codes are being missed, and how each practice compares to specialty peers.
Layer 4 of the Scribara platform — turns encounter outputs into health-system intelligence
Epic FHIRRAPIDSNVIDIA InceptionHEDISHIPAAACO-ready
RAPIDS cohort scan<2 min
Readmission model30-day
Care-gap detectorHEDIS
Peer benchmarkspecialty
The problem
Why Nexar exists
Health systems and specialty groups sit on millions of encounter records but cannot derive actionable population intelligence from them at clinical speed. Existing tools are built on 45-day claims lag, not encounter-quality AI signals.
- GPU-accelerated cohort queries over 100M+ encounter records via RAPIDS [ASPIRATIONAL scale]
- NeMo readmission and progression risk models trained on Scribara's specialty corpus [ASPIRATIONAL]
- HEDIS / MIPS quality-measure computation and value-based care export
- Specialty-peer benchmarking against anonymized cohort
01 / Aggregate
All encounters, one lens.
De-identified Scribara outputs flow into RAPIDS, creating a living encounter data lake queryable in under 2 minutes. [ASPIRATIONAL at 100M+ scale]
02 / Predict
Who needs attention now.
NeMo models score every patient for readmission risk, disease progression, and denial probability — trained on your specialty's own outcome data. [ASPIRATIONAL]
03 / Surface
Where you stand, peer to peer.
Specialty-peer benchmarking compares coding accuracy, prior-auth approval rate, and care-gap closure against the anonymized Scribara cohort.
04 / Act
From insight to intervention.
Flagged cohorts route back to Authra for care-gap outreach and to Forma as outcome-labeled training data. [ASPIRATIONAL]
Capabilities
What Nexar does
Risk prediction
30-day readmission and disease-progression risk scored at every encounter via NeMo predictive models. [ASPIRATIONAL]
Quality reporting
HEDIS, MIPS/MACRA, and value-based care metrics computed directly from encounter evidence.
Peer benchmarking
Compare your specialty group's coding mix, denial rate, and care quality against the anonymized Scribara cohort.
Executive briefings
NeMo LLM generates weekly plain-language population summaries for medical directors and ACO administrators. [ASPIRATIONAL]
Care-gap closure
Flags patients against HEDIS measures and routes recall outreach via Authra. [ASPIRATIONAL]
GPU-essential
Why accelerated compute is required
RAPIDS cohort queries over 100M+ encounter records return in under 2 minutes on GPU; CPU Pandas would take 30–60 minutes. TensorRT-compiled risk models score each patient in under 100 ms — a hard clinical latency requirement. [ASPIRATIONAL at scale]
- RAPIDS (cuDF, cuML) for GPU-parallel cohort analytics
- NeMo for predictive model training and generative briefings
- TensorRT + Triton for <100 ms per-patient risk scoring
- NIM on NVIDIA AI Enterprise for on-prem health-system deploy [ASPIRATIONAL]
RAPIDS cuDF + cuMLcohort
NeMo trainingH100/H200
TensorRT inference<100 ms
Triton servingmulti-tenant
Impact
What it moves
0
Encounter records [ASPIRATIONAL scale]
0
Full cohort scan on GPU
0
Readmission prediction window [ASPIRATIONAL]
0
Population health market (US, 2026)
The ecosystem
Works with the rest of Scribara
Nexar consumes what every other product produces and feeds population insights back.
Answers
Frequently asked
Nexar ingests de-identified encounter outputs from Scribara. PHI is redacted at ingress; per-tenant consent governs any cross-customer analytics.
Yes — NIM packages the analytics and prediction models for VPC or on-prem deployment via NVIDIA AI Enterprise. [ASPIRATIONAL]
Practices are compared against an anonymized cohort of Scribara customers in the same specialty; no individual practice data is ever shared.
See your population, not just your patients.
A focused demo showing Nexar running cohort analysis on a specialty encounter corpus.