03    CASE STUDIES

Selected work,
in production.

Three engagements that crossed the line from research artifact to live clinical deployment. Names, integrations, and metricspublished with client permission.

case 01
Mount Sinai

Health system · Ophthalmology

DICOM
PACS
PyTorch
MONAI
Cutting emergency
retinal wait times by 45%.

End-to-end triage model and clinician interface for a retinal imaging service handling thousands of cases per month.Deployed inside Mount Sinai's network, integrated with their PACS, and validated against radiologist consensus.

45%
Faster triage
0.94
AUROC, reader-validated
12wks
Lab → production
case 02
Maccabi Healthcare

Insurer · Population health

TFLite
Triton
FDA SaMD
Edge
Population-scale risk stratification
across 2.6M members.

Longitudinal cohort modeling for chronic disease progression. We rebuilt their feature store and inference pipeline to runnightly across the full member base, with HL7-driven ingestion and on-prem deployment.

2.6M
Members scored nightly
8 wks
Lab → on-prem
99.9%
Pipeline reliability
case 03
Semler Scientific

Medical device · Cardiovascular

TFLite
Triton
FDA SaMD
Edge
Edge inference on a vascular diagnostic device.

Re-engineered an existing classifier for sub-second inference on a constrained hardware target. Built the eval suite and FDA-ready validation harness alongside the embedded engineering team.

0.96
AUROC on test set
320ms
P99 inference latency
FDA SaMD
Validation track
GET IN TOUCH

Ready to scale your
clinical AI?

Two weeks. A working prototype. No retainer.

A clinical AI engineering practice. We build deep learning systems and production ML for hospitals, pharma, and medical device teams.
© 2026 ReallyGreatTech · Tel Aviv · New York
HIPAA · GDPR · SOC 2 · HL7/FHIR