Health missions · Data science
From claim line to care decision.
Health data is unlike any other: fragmented across standards, governed by layered privacy law, and consequential in a way ad clicks never are. This is the domain we chose — and the discipline it demands shapes everything we build.
HDS · Geo surveillance
TAMPA BAY · WK 30 · PER 100K
Choropleth map of the Tampa Bay region — real Census ZIP-code boundaries for Hillsborough, Pinellas, and southern Pasco counties — shaded light to dark blue by weekly syndromic rate per 100,000. A three-ZIP cluster in East Tampa is flagged in amber and under review; a milder rise shows in south St. Petersburg. Illustrative rates over real geography.
Mission domains
Where we operate
Four problem spaces, each with its own data shapes, statutes, and stakes.
CMS-scale missions
Claims & Payment Integrity
Fraud, waste, and abuse detection across billions of claim lines; risk-adjustment analytics; provider network analysis; and prior-authorization decision support.
Public-health missions
Population Health & Surveillance
Syndromic surveillance pipelines, outbreak detection models, social-determinants integration, and jurisdiction-level dashboards that hold up during a crisis.
VA / DHA-scale missions
Clinical Operations
Risk stratification for care management, capacity and wait-time forecasting, clinical NLP over provider notes, and decision support embedded in the EHR workflow.
NIH-scale missions
Research Data Platforms
OMOP-based research enclaves, registry and trial data systems, de-identification pipelines, and reproducible analysis environments for scientific teams.
Delivery
Health missions, run on the line.
Everything above ships through the Dark Factory™ — our proprietary spec-driven pipelines with programmatic, AI, and human validation at every stage. One example release from each pipeline:
DF/SW
Prior-auth API, in production
A production API shipped through the Dark Factory: twelve signed deploys in thirty days, p95 latency brought under its service-level objective — and every release landed with its accreditation evidence attached.
See the pipelineHDS · DF/SW · Prior-auth API
PROD · 30 DAYS
Uptime · 30d
99.98%
SLO 99.9%
p95 latency
130 ms
SLO 180 ms
Signed deploys
12
all gates green
DF/DA
Claims ingest, governed
Claims ingest released from the Dark Factory: 412 million claim lines in thirty days, with quality gates quarantining bad batches at the door — before they can poison the models and reports downstream.
See the pipelineHDS · DF/DA · Claims ingest
30 DAYS · 42 SOURCES
Claim lines · 30d
412M
+3.4% vs prior 30d
Quality pass rate
99.4%
schema · lineage · privacy
Quarantined
3
batches held at gate
DF/ML
Readmission model, calibrated
A readmission model held to its signed-off spec: calibration a review board can read at a glance, and drift monitoring that triggered an automatic retrain at the gate — before the program ever felt it.
See the pipelineHDS · DF/ML · Readmission model v4.2
RUN 2381 · IN PRODUCTION
AUC
0.81
spec floor 0.78
Calibration error
1.2%
expected ≤ 2%
Last retrain
WK 19
drift gate · auto
DF/AI
Clinical assistant, held to spec
A clinical assistant verified against its behavior spec on every release: compliance meters for what it must do, must refuse, and must escalate — and red-team findings driven to zero before ship.
See the pipelineHDS · DF/AI · Clinical assistant
RELEASE 48 · SPEC v12
Eval cases
1,912
derived from spec
Behavior pass
99.6%
gate floor 98%
Open findings
0
2 releases clean
DF/DX
The console leadership opens
A decision product off the Dark Factory: live feeds, a two-week trend, and jurisdiction rankings — where the one flagged region traces back to governed data and a validated model.
See the pipelineHDS · DF/DX · Executive console
WK 30 · LIVE FEEDS
ED visits · 7d
12.4K
+6.1% vs prior wk
Occupancy
84%
surge floor 90%
Signals open
1
Region 03 · wk 30
DF/YOU
Your mission’s release
A pipeline stood up for your release class — spec kinds, gates, and sign-off authorities defined together.
Fluency
We speak the standards natively.
No translation layer between your data and our engineers.
- FHIR R4
- HL7v2
- C-CDA
- X12 837/835
- OMOP CDM
- USCDI v4
- SNOMED CT
- LOINC
- RxNorm
- ICD-10-CM
Trust architecture
Compliance, engineered in.
The regulatory frame isn't an obstacle to good engineering — it's part of the spec.
- HIPAA
- Privacy and security rules engineered into every data flow, not bolted on.
- FedRAMP
- Cloud architectures aligned to High-baseline controls from the first design review.
- NIST 800-53
- Controls as code with automated evidence collection for continuous ATO.
- Section 508
- WCAG 2.2 AA accessibility tested with assistive technology, every release.
- 42 CFR Part 2
- Substance-use data segmentation and consent management where missions require it.
- IRB & dbGaP
- Research data governance for human-subjects and genomic datasets.