Back to Case Studies
Case study · Healthcare · AWS

AI-Powered Diagnostics Platform Across 120 Hospitals

Pan-India Hospital NetworkAWS7 months18 engineers
40%
Faster radiology turnaround
99.9%
Data availability SLA
120
Hospitals connected
8,000
Daily scans processed
The challenge

Siloed Data, Delayed Diagnoses, Compliance Risk

Radiologists across 120 hospitals were working from disconnected PACS systems, emailing DICOM files between sites, and waiting up to 72 hours for specialist second opinions. Patient data was stored in 14 different formats across on-premises servers with no unified access layer — creating both clinical risk and HIPAA/DISHA compliance exposure.

72-hour average wait for specialist radiology second opinions
14 disconnected PACS systems with no unified data layer
HIPAA and DISHA compliance gaps across on-premises storage
No AI-assisted triage — critical findings buried in scan queues
Our approach

Unified DICOM Lake with AI Triage on AWS

SkySync built a centralised DICOM data lake on AWS S3 with a HIPAA-compliant access layer, then layered a custom AI triage model that flags critical findings within 90 seconds of scan upload. Radiologists across all 120 hospitals now work from a single web-based viewer with real-time collaboration tools.

01

DICOM Ingestion Pipeline

Built AWS DataSync pipelines to ingest from 14 legacy PACS systems into a centralised S3-based DICOM lake with full audit trails.

02

HIPAA-Compliant Access Layer

Implemented AWS PrivateLink, KMS encryption, and fine-grained IAM policies to meet HIPAA and DISHA requirements across all data flows.

03

AI Triage Model Training

Trained a custom CNN on 2.4 million anonymised scans to detect 18 critical finding categories with 94.2% sensitivity, validated by a panel of 12 senior radiologists.

04

Unified Viewer & Collaboration

Deployed a zero-install web-based DICOM viewer with real-time annotation, specialist routing, and mobile-optimised reading for on-call radiologists.

Architecture

HIPAA-Compliant AI on AWS

Ingestion

AWS DataSync + Transfer Family

Automated DICOM ingestion from 14 legacy PACS systems

Storage

Amazon S3 + Glacier

Centralised DICOM lake with lifecycle policies and 7-year retention

AI/ML

Amazon SageMaker

Custom CNN training, hosting, and real-time inference for triage

Compute

AWS ECS Fargate

Serverless containers for the DICOM viewer and API layer

Security

AWS KMS + PrivateLink + Macie

End-to-end encryption, private networking, and PHI detection

Observability

Amazon CloudWatch + X-Ray

Distributed tracing and alerting across all 120 hospital endpoints

Results

Clinical and Operational Impact

40%
Faster radiology turnaround
From 72 hours to under 43 hours average
94.2%
AI triage sensitivity
Validated across 18 critical finding categories
99.9%
Platform availability
Across all 120 hospitals since go-live
Zero
Compliance findings
In first external HIPAA/DISHA audit post-launch
"The AI triage layer has fundamentally changed how our radiologists prioritise their queues. Critical findings that used to wait hours are now flagged in 90 seconds. SkySync delivered something we thought was 3 years away."
Chief Medical Information Officer
Pan-India Hospital Network

Ready to achieve similar results?

Talk to our architects about your cloud and AI challenges. We'll scope a proof-of-value in two weeks.