Structured document understanding
Modeling long, technical documents where the important signal is spread across evidence, requirements, methods, outcomes, tables, and cross-document context.
AI Researcher + Engineer
Reliable AI for structured technical and scientific documents.
I build machine learning systems that turn messy, high-stakes documents into structured, traceable knowledge that people can inspect, validate, and reuse, from clinical research papers to regulated workflows in engineering and law.
Research Themes
From unstructured documents to reusable representations.
My work sits at the intersection of machine learning, NLP, and document understanding. I am especially interested in systems that make reasoning over structured technical and scientific documents more inspectable, evaluable, and useful in operational settings.
Structured outputs and validation
Designing schemas, provenance, and review paths so extracted information can be checked instead of treated as a black-box answer.
Evidence and decision workflows
Building toward workflows that extract numeric and textual signals from complex source material with clearer provenance and reasoning.
Reusable representations for reasoning
Exploring how typed, traceable document representations can support querying, comparison, consistency checks, and downstream reuse.
Selected Publications
Research depth in clinical trial papers and evidence synthesis.
Clinical evidence is the main public research domain where I have published so far. The broader interest is transferable: building systems that extract, structure, validate, and reuse information from complex documents.
Enhancing Study-Level Inference from Clinical Trial Papers via RL-based Numeric Reasoning
A reasoning-driven approach to extracting structured numerical evidence and deriving study-level conclusions for systematic evidence synthesis.
Query-driven Document-level Scientific Evidence Extraction from Biomedical Studies
Introduces CochraneForest and URCA for document-level evidence extraction from full biomedical studies in response to clinical research questions.
AutoForest: Automatically Generating Forest Plots from Biomedical Studies with End-to-End Evidence Extraction and Synthesis
An end-to-end system for suggesting ICO elements, extracting outcome data, performing synthesis, and rendering forest plots from biomedical papers.
Background
Applied AI with a document-systems bias.
I am a Machine Learning Researcher/Engineer at ZAZU Systems, focused on applied AI for complex clinical and document workflows.
I graduated from DCU with an MSc in Artificial Intelligence, specialising in NLP and document synthesis, and was recognised on the Dean's List. I am also a UCD alumnus from my undergraduate studies.
I care most about ML systems that move beyond prototypes: reliable, evaluable tools that help turn complex information into structured knowledge people can actually use.
Elsewhere
A few things outside the documents.
Away from research and engineering, I play jazz, go bouldering, running, and spend time on community volunteering. Those parts of life matter to how I work: practice, attention, shared effort, and showing up consistently.
Contact
Interested in research, applied AI, or document workflows?
I am open to conversations around NLP, evidence synthesis, structured extraction, validation, and practical AI systems for messy technical and scientific documents.