Medical Imaging · Healthcare IT · AWS & AI
Michael Savoy

Michael Savoy

MS, BSRT, CIIP, R.T.(R)(MR)

I started in the scanner room. Now I build the platforms behind it.

I'm a former radiologic technologist who leads medical imaging, clinical data, and cloud work for health systems, imaging vendors, and federal healthcare. I move between DICOM, FHIR, and OMOP, and work across PACS, AWS, Amazon Bedrock, and AI, building the automation that makes it all hold up in real clinical use.

Southern Maryland · Serving DC / MD / VA · Formerly Colorado Springs, CO
See my experience LinkedIn
Reading viewer · CT Chest · Axial ● LIVE
Operator: SAVOY^MICHAEL
CIIP · R.T.(R)(MR)
Southern Maryland
Medical Imaging
Healthcare IT
AWS · Bedrock · AI
W 400  L 40
Slice 128 / 256
Lang: EN · ASL
DICOM · HL7 · FHIR
AI Bedrock · prior ranked 0.97
Experience

Medical imaging, clinical data, and cloud leadership

Health systems, imaging vendors, healthcare AI companies, and federal health, from the scanner room to the cloud.

Lead AWS Cloud Engineer

General Dynamics Information Technology (GDIT) · Federal health
Current

Leads design and build of cloud imaging and AI platforms for federal healthcare on AWS.

  • Multi-tenant radiology reading platform: led the architecture and build, combining per-site edge gateways, AWS HealthImaging, and Amazon Bedrock ranking of relevant prior studies, with hard tenant isolation through ABAC, KMS, and GuardDuty. Isolation was proven in both directions, and each new tenant deploys from the same automation with no manual edits.
  • Demos to federal health stakeholders: took the platform from requirements to a working end-to-end demo.
  • Real-world test data at scale: built Python toolchains to discover, pull, inventory, and analyze open DICOM datasets, including 497K+ instances from NCI Imaging Data Commons.
  • Found an undocumented platform limit: discovered an AWS HealthImaging encryption-key constraint during the build, designed around it, and escalated it to AWS.
AWS HealthImagingAmazon BedrockHealthLakeABAC / KMSDICOMPython

Software Engineer, Cloud & Imaging Informatics

QT Imaging · Breast imaging and clinical research
Recent

Owns the imaging and clinical data flow for a fast-growing scanner network that serves both clinical care and regulated clinical trials.

  • Cloud-native DICOM ingestion pipeline: designed and built an event-driven pipeline to replace manual on-site processes, scaling from 15 to 45 sites. It runs S3 → SQS workers, DICOM Storage Commitment, a dead-letter queue with automated reconciliation, a Postgres audit ledger, and CDK infrastructure as code.
  • PHI remediation at scale: scanned 383,595 DICOM files and flagged 13,599 for PHI, using header checks plus OCR of burned-in pixel text. Built a Flask review tool so the flagged images could be checked visually.
  • Clinical viewer rollout: led the Olea Sphere deployment alongside InteleShare Cloud PACS for about 15 radiologists, most of them remote. Caught and escalated a risk of IRB research data mixing with HIPAA clinical worklists.
  • DICOM conformance and routing: wrote the QTViewer 2.9.0 DICOM Conformance Statement. Built pynetdicom-based forwarding with tag morphing, and resolved association issues with teleradiology and PACS partners.
  • Target-state architecture: designing the imaging and clinical-trial data architecture for a 10–50 site horizon, covering archive, routing, reader, EDC, and trial-imaging layers. Also running a data inventory to move clinical and imaging records out of scattered CSVs into a database.
pynetdicomAWS CDKS3 / SQSPostgreSQLInteleShareOleaOCR

Senior Data Engineer, Technical Lead

ConcertAI · Oncology real-world data and AI
  • Enterprise imaging pipelines: led pipelines processing millions of images for oncology data and AI model development, including GPU and high-performance compute workloads and work with pharma and biotech clients.
  • Multi-vendor DICOM migration: automated migration of studies from many PACS vendors into centralized TeraRecon archives using Python, PowerShell, and AWS.
  • Quality at scale: built QA and visualization dashboards, and applied ML object detection and forecasting to catch data-quality problems early.
DICOMAWSGPU computeTeraReconPython

Informatics Engineer, R&D

Imidex · Imaging AI
  • Led the pulmonary embolism detection model: owned the object detection work from data curation through training.
  • Data for AI: built tools to convert non-standard image formats into medical-grade DICOM, plus curation and training pipelines in Python and PySpark on Google Cloud.
  • Clinical decision support: built Plotly dashboards for the clinical teams.
Object detectionTensorFlow / KerasPySparkGCPDICOM

Technical Specialist L2, Vue PACS

Philips Healthcare · North America
  • Enterprise PACS support: Tier 2 escalation for Philips Vue PACS customers across North America. Partnered with engineering on complex defects and patches.
  • Interoperability: strengthened DICOM and HL7 integrations, and maintained VMware and AWS storage environments.
Vue PACSDICOMHL7VMware

Senior Systems Engineer

Medisys Health Network
  • Epic integration: optimized integration between Epic and ancillary systems, and wrote automation that reduced operational overhead.
  • Monitoring: implemented cloud-based server monitoring and improved DICOM workflow efficiency.
EpicAutomationMonitoring

Technical Lead, PACS & Imaging Informatics

Montefiore Medical Center · Nearly a decade
  • First cloud PACS: led implementation of Montefiore's first AWS-based vascular PACS.
  • NIH-funded research: engineered mobile ultrasound transmission workflows supporting one of the largest NIH-funded cardiovascular studies.
  • Reliability: designed failover architectures and automated server monitoring for clinical imaging.
PACSAWSUltrasoundDICOM

PACS Administrator

MedStar Harbor Hospital
  • Turnaround: cut radiology report turnaround from two days to four hours by redesigning the dictation workflow.
  • Full PACS overhaul: ran radiology, cardiology, and fetal assessment informatics, and led an upgrade of hardware, interfaces, and software.
PACSHL7Dictation
Interoperability

Fluent across DICOM, FHIR, and OMOP

Imaging, clinical, and research data rarely live in the same standard. I work in all three, and I build the bridges between them so a scan, a patient record, and a research cohort can all describe the same patient.

DICOM
Imaging
  • DIMSE and DICOMweb, C-STORE, Q/R, Storage Commitment
  • Conformance statements and SOP classes
  • Routing, tag morphing, de-identification
  • AWS HealthImaging, PACS, VNA
FHIR
Clinical exchange
  • ImagingStudy, DiagnosticReport, Patient, Encounter
  • HL7 v2 feeds into FHIR resources
  • AWS HealthLake
  • EHR data handling for analytics
OMOP
Research & analytics
  • OMOP CDM ETL from EHR and FHIR sources
  • Standard vocabularies and concept mapping
  • Imaging metadata into research-ready cohorts
  • Real-world data for AI and outcomes work
DICOM → FHIR

Turn study and series metadata into ImagingStudy resources, so images show up in the clinical record.

FHIR / EHR → OMOP

Map EHR data, including FHIR-shaped sources, into OMOP tables and standard concepts for research and analytics.

DICOM → OMOP

Bring imaging into OMOP-based research so cohorts can be built across both images and clinical history.

Builds

Things I've built on my own

Tools I design and write myself, because building is how I learn.

Densflow

A config-driven Python ETL framework for DICOM, CSV, and SQL sources. YAML configs and Pydantic models drive ingestion, transforms, and validation, with Airflow DAGs and PostgreSQL behind it.

Local RAG voice assistant

A retrieval-augmented assistant over a library of about 20,000 articles, running fully on local hardware with LLaMA, Qdrant, Whisper speech-to-text, and ElevenLabs voice.

On-call phone automation

A Python and Twilio on-call phone system for a volunteer hospital liaison team, handling call routing and escalation.

How I work

I speak both clinical and engineering

Most imaging projects that fail have the right technology. They fail because the people running the scanners and the people building the systems never fully understood each other.

01

Start in the reading room

I've run the scanners and supported the radiologists. Every design starts with the technologist's and the reader's workflow, not the architecture diagram.

02

Prove it with something that runs

A working demo on real DICOM data answers questions a slide can't. I build early, show early, and let the results carry the conversation.

03

Build for whoever supports it next

Automation, monitoring, and clean handoffs come standard. A system is only done when the team that inherits it can run it without me.

Fluent in American Sign Language

I communicate directly with Deaf and hard-of-hearing patients and colleagues. It has shaped how I think about accessible communication, both in clinical settings and on the teams I lead.

Expertise

Where I go deep

Medical imaging & PACS

PACS and VNA, modality integration, Epic Radiant, and radiology workflow.

DICOM & interoperability

Conformance statements, routing, PHI de-identification, HL7 v2, and large-scale migration.

AWS for healthcare

AWS HealthImaging, HealthLake, S3, IAM and ABAC, KMS, and secure multi-tenant design.

Healthcare AI & Bedrock

Amazon Bedrock, agentic AI, LLM extraction from clinical text, and ML for imaging data quality.

Clinical data & OMOP

OMOP CDM ETL and vocabularies, FHIR, clinical-trial imaging data flow, and research pipelines.

Automation & app development

Python tooling, dashboards, monitoring, and custom healthcare applications.

Credentials

Certifications and education

Credentials

Certified Imaging Informatics Professional (CIIP)
AWS Certified Solutions Architect – Associate
NVIDIA-Certified Professional: Agentic AI
CompTIA Security+
ARRT Radiography (R) & MRI (MR)
Epic Radiant & Cupid

Education

M.S., Health Information Systems
CUNY School of Professional Studies

B.S., Radiologic Sciences
CUNY New York City College of Technology

Radiography
Johns Hopkins Schools of Medical Imaging

Mission Clinical IT

Independent consulting

Through Mission Clinical IT LLC, my Maryland consulting firm, I take on a limited number of advisory engagements in medical imaging and clinical data.

  • Imaging and clinical-trial data architecture
  • PACS and DICOM workflow assessment
  • AWS and AI for imaging
Beyond work

Outside the reading room

Aquascaping Poison dart frogs Landscape photography Hiking Snowboarding 3D printing

Let's connect

I'm always glad to talk with people working on medical imaging, clinical data, or AI in healthcare, especially in federal health and the DC / Maryland / Virginia region.

Email me Connect on LinkedIn