Jason Thomas, PhD
Healthcare Data Operations & AI/MLOps Platform Leader | Multimodal Medical Data & Partnerships
Philips Image Guided Therapy Devices
Biography
I lead healthcare data operations, regulated AI/MLOps platform strategy, and data governance for medical-device R&D. My work sits between engineering, clinical informatics, product AI, regulatory evidence, privacy, and business strategy: translating product roadmaps into data demand/supply models, acquisition and annotation priorities, fit-for-purpose criteria, platform roadmaps, and reusable assets for development, validation, analytics, and real-world evidence.
At Philips Image Guided Therapy Devices, I built and now lead a distributed Data & AI Platform Engineering function spanning data sourcing, curation, annotation, data/ML engineering, infrastructure, MLOps, stewardship, governance, and real-world data. I manage a multi-million-dollar platform budget, support more than 100 developers and analysts, and help shape healthcare data and AI roadmaps, investment logic, success metrics, and executive dashboards across senior R&D and business-unit leadership.
My background also spans biomedical informatics research, real-world evidence, synthetic-data validation, medical NLP, ECG/VCG/ECGi, speech/audio biomarkers, clinical imaging context, and intraprocedural multimodal data. Over time, I am especially interested in the data, validation, and platform foundations required for physical AI in healthcare, including AI-assisted procedures, medical robotics, simulation, and sensor-fusion workflows.
- Healthcare data operations and strategy
- Regulated AI/MLOps platforms
- Multimodal medical and sensor data
- Data partnerships, licensing, and governance
- Real-world evidence and clinical informatics
- Privacy-preserving data utility and validation
- Medical data standards and interoperability
- High-performance technical team building
-
PhD Biomedical Informatics; Data Science Specialization
University of Washington - Seattle WA
-
Bachelor of Science - Human Physiology | Biology (Chem minor)
University of Oregon - Eugene OR
Skills
Roadmap-driven data demand/supply modeling, fit-for-purpose criteria, acquisition and annotation portfolios, investment metrics, and release planning.
Governed data platforms for AI development, validation, analytics, and real-world evidence.
Imaging context, EHR/claims, ECG/VCG/ECGi, audit logs, speech/audio biomarkers, and intraprocedural data.
Data-use agreements, licensing decisions, active learning and annotation operations, de-identification, provenance, and access controls.
Data lakes/lakehouses, data contracts, lineage, dataset versioning, quality gates, and reproducible pipelines.
HIPAA, GDPR, IRB workflows, EU MDR, EU AI Act readiness, QMS-aligned controls, and regulated evidence.
Computable phenotypes, clinical features, post-market surveillance, real-world evidence, and PMCF analytics.
OMOP, HL7 FHIR, SNOMED, LOINC, ICD-10, CPT, HCPCS, RxNorm, UMLS, and medical data harmonization.
Python, SQL, AWS, Airflow, PySpark, Parquet, Redshift, Athena, SageMaker, ClearML, GitHub Actions, and Linux.
Hiring, mentoring, platform operating models, technical roadmaps, budget ownership, and cross-functional leadership.
Experience
- Own healthcare data operations and AI/MLOps platform strategy for a global medtech R&D organization, translating product roadmaps into data demand/supply models, acquisition and annotation priorities, fit-for-purpose criteria, governance controls, platform roadmaps, investment metrics, and executive dashboards.
- Built and now lead a distributed 0-to-8 platform function across data sourcing, curation, annotation, data/ML engineering, infrastructure, MLOps, stewardship, governance, and clinical informatics.
- Manage a multi-million-dollar platform budget and support 100+ developers, analysts, and product AI/algorithm teams through governed self-service data access.
- Set portfolio priorities for healthcare data partnerships, licensed data, labeled and unlabeled data needs, vendor/tooling choices, annotation operations, lakehouse architecture, and AI-ready dataset reuse.
- Architect platform patterns for DICOM-linked clinical/imaging data, real-world evidence, and intraprocedural multimodal medical/sensor data used in AI-assisted workflows.
- Establish regulated AI validation foundations including lineage, reproducibility, dataset/model versioning practices, quality gates, train/validation/test governance, and blinded held-out evaluation sets.
- Designed and scaled real-world-evidence analytics across more than 1,000 hospitals and 1B+ patient visits for post-market surveillance, regulatory evidence, clinical strategy, and commercial decision support.
- Standardized a reusable library of 150+ computable phenotypes and clinical features used across regulated submissions, analytics, and downstream AI workflows.
- Reduced dashboard build time from roughly two days to roughly ten minutes through API automation and reusable evidence-generation patterns.
- Delivered production AI/GenAI systems on AWS, including identity-resolution and graph/vector retrieval workflows with evaluation harnesses for precision, recall, F1, and regression testing.
- Helped author data and AI strategy narratives, governance practices, and roadmap proposals adopted by senior leadership.
- Built OMOP-based data warehouse and pipeline assets; assessed whether real and synthetic EHR/audit-log data were fit for secondary use under privacy-preserving constraints.
- Contributed to National COVID Cohort Collaborative synthetic-data validation work and analyzed more than 1.8M SARS-CoV-2 tests for geospatial and temporal epidemiologic utility.
- Developed multimodal ML models combining speech/audio, language, and clinical history to predict dementia status and evaluate clinical text/voice biomarkers.
- Trained GPU-accelerated ML models on Slurm-managed HPC infrastructure and AWS.
- Released reusable public healthcare data artifacts, including medical NLP code/data for replication and a 2,882-question annotated medical question-type dataset.
- Developed predictive models and annotation workflows across 100k+ ECGs; sourced clinical data under HIPAA and mined EHR outcomes for publication-ready analyses.
- Collected and analyzed multimodal electrophysiology and imaging-linked data including ECGs, 128-electrode body-surface ECG/ECGi recordings, device interrogations, and intracardiac EGMs during cath-lab procedures.
- Co-authored work combining CT/MRI/3D geometry and electrode localization for cardiac activation mapping.
- Recruited 350+ participants across observational and interventional studies, wrote study protocols, and supported multiple IRB approvals.
- Cofounder 2013, Executive Director 2014+. 300-person race raising $ for rural health EMS
- Total responsibility for P&L, logistics, legal compliance, 5-10 person team
- Live radio interviews
- Worked at >15 different clinics performing electrocardiograms, blood draws, rapid tests, in-person scheduling & billing, training of >10 new employees and process improvements
- Conducted & recorded results of V02 max exercise tests and altitude chamber studies with human subjects, processed lab specimens, subject recruiting and scheduling, data analysis
- Managed ~10 direct reports per shift in a 250k ft2 facility, first responder & responsible for safety of all students & staff, developed new hiring process to screen 700 applicants
- Aided healthcare delivery to a mainly low-income and special needs population
- Assembled outgoing prescriptions for individual patients and kept track of inventory
- Assisted in clinic-wide conversion of paper medical records to digital files
