MD Anderson Cancer Center | Clinical Research Ontology Automation
The Challenge
The existing toolset relied on manual curation of ontology content. Users needed a faster, more repeatable process that could scale laterally to multiple clinical domains and integrate easily with their internal terminology handling systems.
Strategic Approach
To demonstrate feasibility and accelerate innovation within a compressed timeline, I established the technical vision and foundational architecture for a 12-factor cloud-native system. This involved defining and integrating 24 services, including microservices, micro-UIs, and serverless functions. I led the technical design and proof-of-concept implementation, integrating AI-augmented development (e.g., Gemini 2.5 Pro for medical term normalization) to streamline complex tasks. The architecture was designed to support the full ontology lifecycle using OWL API, SNOMED CT, RxNorm, and UMLS, and was deployed on-premises within their regulated data center, demonstrating compatibility with Protégé and incremental reasoners.
Impact & Results
- 100% accuracy in validated outputs
- 85% reduction in manual QA effort
- Production-viable ontology tooling ecosystem delivered within 16-week POC timeline
- Validated repeatable pattern for integrating AI-assisted automation into regulated clinical research environments