My 2015 PhD in Computer Science at the University of Colorado Boulder investigated drug–drug interactions through statistical inference and Semantic Web technologies. I build on that biomedical foundation today to study the evidence, judgment and responsibility behind AI-assisted decisions.
2015PhD in Computer Science, University of Colorado Boulder
D3Pharmacovigilance system for characterizing drug–drug interactions
Big DataStatistical inference across heterogeneous biomedical evidence
Semantic WebKnowledge representation as part of decision support
The dissertation
Characterization of Drug-Drug Interactions through Statistical Inference and Semantic Web Technologies developed a computational approach to pharmacovigilance that combined inference from large-scale data with structured biomedical knowledge.
The work was not only about detecting possible interactions. It was about the harder systems problem: how incomplete, heterogeneous and differently represented evidence changes what a computational system can safely conclude.
The enduring research question became: what does a system hide when its summary appears correct?
The doctoral research sat at the intersection of natural-language and computational methods, biomedical informatics, linked data and pharmacovigilance.
ADVISOR · 01
James H. Martin
PhD Advisor · University of Colorado Boulder. His research group page documents the doctoral lineage and lists the 2015 dissertation on drug–drug interaction characterization.
Contribution: computational rigor, language and data-driven inference
Context: Computer Science, University of Colorado Boulder
Principal Co-Supervisor · Domain Expert. Biomedical informatics and Semantic Web expertise shaped the knowledge-representation and translational dimensions of the doctoral work.
Contribution: biomedical knowledge graphs and semantic technologies
Continuity: drug discovery, personalized medicine and FAIR data
Later projects changed domains, but the intellectual structure stayed recognizable: examine what the representation omits, test whether aggregate results hide meaningful differences, and connect technical evaluation to the decision that follows.
01 · KNOWLEDGE
Completeness matters
Missing or uneven knowledge changes the apparent reliability of a system.
02 · REPRESENTATION
Structure matters
How evidence is encoded changes what can be inferred and connected.
03 · EVALUATION
Summaries can hide structure
Aggregate performance may conceal differences that matter at the decision level.
04 · ACTION
Decisions are the endpoint
The system is not finished when it produces a score; the consequential question is what happens next.
From D3 to today
A continuous research arc.
D3 and D4 connect biomedical evidence to mechanistic inference. My current AI ethics agenda develops that concern into questions of justice, accountability and human agency, with RIDI as one current strand.
D3 / D4Connecting biomedical knowledge, prediction and mechanistic explanation.
GenomeFitGenomic evidence translated toward personalized screening and drug response.
RIDIWhat aggregate AI evaluation hides about allocation identity.
Institutional SystemsHow evidence becomes policy, operating models and consequential decisions.
Interested in this research lineage or a related collaboration?
Current work spans decision intelligence, biomedical knowledge, digital health and evaluation of consequential systems.