Biomedical AI Researcher | Clinical NLP, LLM Evaluation & Precision Medicine
Hi, I’m Cindy.
I’m a PhD candidate in Bioinformatics at the University of British Columbia (UBC), co-supervised by Prof. Wyeth W. Wasserman and Prof. Jian Zhu.
My background brings together healthcare, pharmacogenomics, biomedical AI, and machine learning. I hold an MSc in Bioinformatics, a Doctor of Pharmacy (PharmD), and a BSc in Pharmaceutical Sciences. Before moving into research, I practiced as a registered pharmacist, providing direct patient care.
My research focuses on large language models (LLMs), biomedical natural language processing, and auditable AI systems for rare-disease diagnosis. I develop LLM-based pipelines for extracting and structuring clinical evidence from unstructured narratives, alongside embedding-based methods for phenotype representation and diagnostic retrieval. My work spans representation learning, dense retrieval, and ontology-based reasoning, with particular attention to model capabilities, failure modes, and the traceability of predictions to supporting clinical evidence.
Drawing on my clinical background and training in bioinformatics, I aim to develop AI systems that connect clinical evidence with biomedical research to improve patient care and quality of life.
Evaluates ontology-defined relationships and joint phenotype constraints, identifying limitations in relational and compositional retrieval that concept-grounding performance alone does not reveal.
Accepted to the Conference on Empirical Methods in Natural Language Processing, 2026.
Combines pathway-based genetic analysis and machine learning to investigate predictors of medication toxicity and the predictive contribution of pathway-informed features.
Clinical Pharmacology & Therapeutics, 119(2), 447–456. Online November 2025; issue February 2026.
Introduces an expert-curated benchmark for structured extraction from rare-disease case reports, emphasizing inherited metabolic diseases and clinically relevant extraction errors.
Conference on Health, Inference, and Learning. PMLR 287:527–542, 2025.
Introduces a structured dataset of 85,961 medical case reports and examines the gap between lexical similarity and clinical relevance in embedding-based retrieval.
24th Workshop on Biomedical Language Processing, ACL 2025, pp. 249–262.
Testing whether retrieval respects biomedical knowledge.
I designed a benchmark spanning concept grounding, ontology relationships, and joint phenotype constraints to examine where dense retrieval captures clinical relevance—and where it fails.
Turning published case reports into structured evidence.
I built LLM-based clinical NLP pipelines integrating structured extraction, phenotype representations, embedding-based retrieval, and Neo4j knowledge graphs. The companion CaseReportBench evaluates extraction against expert-curated clinical evidence.
Connecting candidate diagnoses to patient evidence.
I developed a diagnostic-support research pipeline for de-identified specialist consultation notes, combining multi-stage evidence extraction, deterministic and semantic validation, candidate disease ranking, and feature-level explanations for clinician review.
Traceability from a candidate diagnosis
to the clinical findings that support it.
Pathway-informed learning for pediatric oncology.
I led a study integrating clinical and genetic data to investigate methotrexate-induced mucositis. I developed and validated XGBoost models using feature ablation, calibration assessment, and SHAP interpretation to evaluate the contribution of genomic features.
Python · PyTorch · Hugging Face · Neo4j · XGBoost · Docker / Apptainer · Slurm
Evaluation work includes explicit baselines, tie-aware ranking metrics, cluster-bootstrap confidence intervals, perturbation tests, and attribution-stability analyses.
Jan 2024 – Present
Centre for Molecular Medicine and Therapeutics, UBC
Biomedical NLP, phenotype representations, ontology-aware retrieval, and auditable diagnostic-support research. Collaboration with clinicians, geneticists, computer scientists, and ontology researchers.
Jan 2022 – Dec 2023
Canadian Pharmacogenomics Network for Drug Safety, UBC
Pathway-informed machine learning for medication safety, integrating clinical covariates and SNP-level data covering 1,036 genes and 18 biological pathways.
Clinical experience
Pharmacy practice · British Columbia, Canada
Direct patient care through medication management, clinical assessment, patient counselling, and collaboration with healthcare professionals.
For conversations about biomedical NLP, scientific evidence intelligence, or AI research in biotechnology and pharmaceutical R&D, connect with me on LinkedIn. You can also find my publications on Google Scholar and my code on GitHub.
2023–2025
Supported graduate-level courses in programming, algorithms, statistical inference, data visualization, and cloud computing through tutorials, laboratory instruction, technical guidance, and project mentorship.
2021–Present
Facilitated case-based learning and clinical skills training for PharmD students and international pharmacy graduates, focusing on pharmacotherapy, medication management, clinical reasoning, and patient safety.
2020–Present
Mentored undergraduate students in biomedical NLP, clinical AI evaluation, and scientific communication. Guided research design, experimental evaluation, and interpretation of model outputs, contributing to conference publications and research presentations.